View original document

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

S E P T E M B E R /O C T O B E R 1996

Bruce J. Summers is a senior vice president and chief financial officer at the Federal Reserve Bank of Richmond. R. Alton Gilbert is a vice
president and banking advisor at the Federal Reserve Bank of St. Louis. Mary C. Lohmann provided research assistance.

began declining in 1980, when the Reserve
Banks began charging for payment services, as required by the Monetary Control
Act. This declining share, for both smalland large-dollar payments, appears to represent a major shift in the operation of the
U.S. dollar payments system. Our paper
examines the implications of this shift for
the Federal Reserve’s ability to fulfill its
mandate to safeguard the stability and
efficiency of the payments system. In
particular, we examine whether the problems that existed in the payments system
prior to 1914 will at some time reappear as
the Fed’s operational role declines. It is
important to consider whether the nation’s
payments system has changed in ways that
make Reserve Bank services less essential
for dealing with the problems that have
beset it in times past, and what the future
role of the Federal Reserve Banks should
be as payment processing systems continue to evolve.
The following section examines the
operation of the payments system prior
to the formation of the Federal Reserve,
focusing on aspects of the system that
were considered defects by advocates of a
central bank. Subsequent sections establish a conceptual framework for our analysis and describe the payment services
offered by the Reserve Banks and trends
in their share of the total volume and
value of U.S. dollar payments processed
each year. The article then discusses reasons for the declining Reserve Bank share
of payment processing and the implications of these trends for the payments
system.

Clearing and
Settlement of
U.S. Dollar
Payments: Back
to the Future?
Bruce J. Summers and
R. Alton Gilbert
The Federal Reserve System was
formed in 1914. Wide dissatisfaction with
routinely expensive and slow settlement of
interregional payments, as well as occasional disruptions of the payments system
caused by banking panics, are among the
factors that led to its creation. Accordingly,
an important purpose for creating the Federal Reserve to serve as the nation’s central
bank was to enhance the efficiency and
improve the stability of the nation’s payments system.
At the time of the formation of the
Federal Reserve, the paper check was the
principle means of making payment. The
Federal Reserve attempted to fulfill its
mandate for improving the check-collection system by providing banks with a
national check-collection service.1 Since
it was the only institution with a nationwide network of banking offices and
settlement accounts for banks, it had an
advantage in interregional check collection. Over time, the Reserve Banks added
new payment services to exploit the advantages of new technology: wire transfer
of reserves, a book-entry service for safekeeping and electronically transferring
ownership of government securities, and
the automated clearinghouse, designed as
an electronic alternative to checks.
The share of U.S. dollar payments processed through the Federal Reserve Banks

PROBLEMS WITH THE
PAYMENTS SYSTEM PRIOR
TO THE FORMATION
OF THE FEDERAL RESERVE
An analysis of the importance of
Reserve Bank payment services for the
banking industry in the United States

FEDERAL RESERVE BANK

3

OF

S T. L O U I S

1

For convenience, depository
institutions are called banks.

S E P T E M B E R /O C T O B E R 1996

2

The discussion of the checkcollection system prior to 1914
and changes made by the
Federal Reserve is based on
Spahr (1926), chapters IV, VI,
and VII; Watkins (1929),
chapter VI; and White (1983),
chapter 2.

requires a review of some banking history.2
By the mid-1850s, the dollar value of U.S.
bank deposits exceeded that of banknotes,
and the value of transactions settled by
check exceeded the value of transactions
settled by banknote. This growth in
check transactions required a system for
clearing a large number of checks among
banks. Before the introduction of Federal
Reserve services, commercial banks
cleared checks drawn upon other local
banks by channeling them through local
clearinghouses or delivering them directly
to the local banks for payment. Typically,
local checks could be collected quickly
and at par.
Collecting checks drawn on banks
outside a given community involved more
time and expense. When checks were presented directly to paying banks at their
place of business, the banks were required
by law to pay the face value of the checks.
Banking law did, however, permit banks to
pay less than the face amount of checks
submitted for collection by indirect means,
such as through the mail. The rationale
for this deduction from the face amount,
called an exchange charge, was that remitting payment could involve certain costs,
including the cost of transporting coin or
bank notes from the paying bank to the
collecting bank. Delays were another
expense to collecting banks, in addition
to exchange charges. Under banking law,
a bank that received checks through the
mail became the collecting agent for the
bank that had sent them and was therefore
responsible for obtaining payment from
itself. As a result, paying banks often
remitted funds to collecting banks several
days after receiving checks through the
mail.
Despite the rationale for exchange
charges, many bankers considered them
a basic defect in the operation of the payments system. Prior to the formation of the
Federal Reserve System, there were several
major proposals and attempts by bankers
to eliminate exchange charges. Opposition
to exchange charges was most common
among bankers in the larger cities, where
banks generally paid for checks drawn

on accounts of their depositors at par,
through local clearinghouses. The banks
that imposed exchange charges generally
were relatively small and located in more
isolated areas.
Collecting banks attempted to avoid
these delays and exchange charges by
using the services of correspondent banks.
Often depository banks (the banks of
first deposit) sent checks drawn on banks
outside their communities to their correspondent banks. The correspondents
would then send the checks to other banks
with offices near the paying banks, which,
in turn, would present the checks to the
paying banks over the counter. In this
system of collection through correspondents, depository banks might receive less
than the face amount of the checks, but
more than if the checks were sent directly
to paying banks. The correspondents
would split the collection fee (the difference between the face value of the checks
and the amount credited to the demand
accounts of the depository banks) with the
other banks that had assisted them in getting the checks to the paying banks. In
some arrangements, the correspondents
would credit the demand accounts of
depository banks for the full amount of the
checks being collected but would require
the depository banks to hold large demand
balances as a form of compensation for
this service. Under either arrangement,
it was competition among correspondent
banks that tended to reduce the costs of
collecting interregional checks.
The process of collecting checks
through correspondents as a means of
avoiding exchange charges led to some
notorious cases of checks passing through
the offices of many banks and traveling
over very long distances, relative to the
actual distance between the depository
bank and the paying bank. Many of the
resulting delays and operating expenses
could have been avoided through more
direct collection channels. Competition
among correspondent banks, however,
led to substantially reduced levels of
exchange charges over time (Spahr, 1926,
pp. 102–3).

FEDERAL RESERVE BANK

4

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Although exchange charges declined
substantially over time, many bankers
continued to view them as a fundamental
defect in the operation of the nation’s payments system. Congress responded to calls
for reform by giving the Federal Reserve a
mandate to improve the efficiency of the
payments system, and the Federal Reserve
responded by establishing a national network of offices for collecting checks. Because the Federal Reserve Act forbids the
Reserve Banks from paying exchange
charges to banks, the Reserve Banks established the practice of accepting for deposit
only those checks drawn upon banks that
had agreed to pay the Reserve Banks at
par.3 Although the Federal Reserve was not
granted legal authority over the exchange
charges set by individual banks, its dominant operational role in check collection
eventually made collection at par (zero
exchange charges) a national standard for
the banking industry.
Another problem with the operation of
the payments system prior to the formation of the Federal Reserve in 1914 was
the occasional disruption of the payments
system caused by banking panics. When
events caused depositors to lose confidence in the safety of their deposits, they
demanded payment in gold coin, banknotes or greenbacks.4 Banks located outside the major financial centers maintained
large shares of their cash assets in deposits
with major banks in the financial centers,
particularly New York City, and they
tended to respond to depositors’ substantial cash withdrawals by drawing down
deposits with these banks. Sprague (1910),
in his analysis of banking panics in the
national banking era, emphasized that the
concentration of bankers’ deposits in a
small number of banks in New York City
made the banking system vulnerable to
disruption.
Bankers attempted to cope with panics
through cooperative arrangements implemented through their local clearinghouses.
During normal times, activities of the
clearinghouses were limited largely to
check clearing and settlement: Banks
deposited gold with the clearinghouses

and received certificates that served as
claims on the gold; they cleared checks
through the clearinghouses and settled
their net positions with clearinghouse
certificates. At times of relatively high
depositor demand for gold and currency
(banknotes and greenbacks), the clearinghouses created additional certificates for
interbank settlement, called loan certificates. Banks that borrowed these additional certificates from their clearinghouse
pledged some of their commercial loans or
other securities to the clearinghouse as
collateral. This process of accepting bank
loans as collateral and issuing loan certificates had the effect of increasing the
monetary base. Members of the clearinghouse could use the gold and currency in
their vaults to meet the demand of their
depositors without concern that they
would have insufficient cash assets to
cover net debit positions at the clearinghouse.
On several occasions after clearinghouses had created loan certificates for
their members, clearinghouse members
also suspended currency payments to their
depositors. While creation of loan certificates helped banks respond to unusually
large demands for currency, the loan
certificates were used primarily to settle
interbank positions with the clearinghouse. Banks were obligated to pay their
depositors gold or currency but did not
always do so when their inventories were
inadequate to meet the demand of their
depositors.5 Instead, during some general
suspensions of currency payments to
depositors, banks paid their depositors
small-denomination loan certificates,
issued by their clearinghouses, which
served as substitutes for currency in emergency situations (Andrew, 1908).
Before the creation of the Fed, when
banks in major financial centers suspended
currency payments to depositors, major
disruptions in the payments system resulted. There is evidence that these suspensions, each of which lasted only one
or two months during the period from
the Civil War through 1914, seriously
disrupted economic activity, including

FEDERAL RESERVE BANK

5

OF

S T. L O U I S

3

Under the Federal Reserve’s
Regulation J, which governs the
collection of checks and other
items by Reserve Banks, an
“item” does not include a
check that cannot be collected
at par. Further, the Reserve
Banks are required to accept
cash and other items at par.

4

After passage of federal banking legislation in the 1860s,
the paper currency in circulation
comprised notes issued by
national banks and greenbacks
(fiat currency issued by the
United States Treasury).
Because national banks were
required to back their notes
with U.S. Treasury securities
deposited with the Treasury
Department, the public considered national bank notes as
safe as gold coins, even during
banking panics.

5

Some have criticized the
national banks in New York City
for suspending currency payments when they still had large
amounts of gold and currency
in their vaults. These critics
maintain that the banks were
too concerned about meeting
their legal reserve requirement—vault cash (gold and
currency) that equalled or
exceeded 25 percent of their
deposits—rather than using all
of the cash in their vaults to
meet demands of their depositors (Dewald, 1972).

S E P T E M B E R /O C T O B E R 1996

interregional and foreign trade, particularly in 1873 (Sprague, 1910, pp. 71–82)
and in 1893 (pp. 119–210).
The Federal Reserve System, modeled
after the clearinghouses of the period, was
authorized to deal with panics by increasing bank reserves through discount window loans.6 Its creators assumed that
government sanction would lessen the
impact of banking panics, and an experience in 1914, just before the Fed was
created, may support this assumption. The
outbreak of war in Europe triggered runs
on U.S. banks. However, the AldrichVreeland Act of 1908 had authorized clearinghouses to put into circulation emergency
issues of national banknotes, which had
been printed and stored for such an event.
Roberds (1995), who finds that the real
economic impact of the panic of 1914 was
smaller than that of prior panics, argues
that the difference can be attributed to
government sanction for the emergency
issuance of national banknotes.

involves discharging the payment obligations. To illustrate the distinction between
these two functions, consider the clearing
and settlement of checks among banks
that are members of a clearinghouse.
Banks rely on the clearinghouse to perform
the clearing function when they exchange
checks drawn on each other. Then the
clearinghouse calculates the multilaterally
netted payment obligations due to and due
from each clearinghouse participant. Banks
participating in the clearinghouse have
various options for settling these obligations. Members of the clearinghouse can
agree to settle using cash or more likely the
deposit liabilities of a private bank, which
might also be a member of the clearinghouse, or through another institution.
Alternatively, settlement could be accomplished through the transfer of reserves
maintained at the Reserve Banks. Using
Federal Reserve Bank liabilities to achieve
interbank settlement is important from a
public policy perspective for at least two
reasons: First, reliance on Reserve Bank
liabilities contributes to the robustness of
settlement arrangements and reduces the
moral hazard that might result if all providers of payment services relied upon a
small number of large commercial banks as
settlement intermediaries. Second, it is by
offering Fedwire and net settlement services
to clearinghouses that the Federal Reserve
is able to exert an indirect form of supervisory influence on the safety and soundness
of private clearing arrangements, since the
Federal Reserve lacks statutory authority
over the operations of clearinghouses. (See
Juncker, Summers, and Young, 1991.)

CONCEPTUAL FRAMEWORK
This section describes some concepts
that are fundamental to understanding the
operation of the payments system and the
role of the Reserve Banks.

Clearing and Settlement

6

See Dwyer and Gilbert (1989)
and Roberds (1995). During
the 1930s, the Federal Reserve
was not effective in dealing
with banking panics. One view
is that banks relied on the
Federal Reserve to deal with
the panics, and the Fed did not
fulfill its role as the central
bank in the face of bank runs.

In describing Reserve Bank services,
it is useful to distinguish between two
processes: the clearing of payments and
interbank settlement of payment obligations. Most Reserve Bank services
combine the clearing and settlement
functions, although the Reserve Banks
also offer interbank settlement services,
with the clearing of payments among the
banks performed through private channels. The implications for the payments
system of declining Reserve Bank operations depend on which function of the
Fed is affected more: clearing or settlement.
Clearing comprises three main steps:
processing payment instruments, delivering them to paying banks, and calculating
interbank payment obligations. Settlement

Network Effects
According to the literature on industrial organization, an industry has network
effects if the value of a service to a customer depends on the number of other
customers using the service. These network effects have important implications
for industry structure and competitive
behavior (see Economides and White,
1994; Katz and Shapiro, 1994).
Because the payments system has
some of the characteristics of a network

FEDERAL RESERVE BANK

6

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

One alternative, of course, to using
the Fed’s network is to develop a private
network for interregional check clearing.
Developing such an alternative would have
been especially difficult prior to 1980,
however, when the Reserve Banks provided payment services to member banks
only, free of explicit charge. Prior to passage of the Monetary Control Act of 1980,
which required the Reserve Banks to
charge for their services, almost all of the
banks that cleared a high volume of interregional checks were members of the
Federal Reserve System. A rival to the Fed
for interregional check clearing would
have had to convince banks to pay a positive charge per check (compared to a zero
charge per item in the Fed’s system) or
withdraw from Fed membership and
rely on the new private system. Once the
Reserve Banks began assessing checkclearing charges and requiring all banks
to maintain reserves, the private systems
for check clearing became more viable
alternatives to the clearinghouse services
of the Reserve Banks.
Industries with strong network effects
also tend to be highly concentrated. If private payments networks were to supplant
the role of the Reserve Banks, this development would raise antitrust issues with
respect to access to the payments system.
Thus, the declining role of the Reserve
Banks in processing payments and the
development of private systems for check
clearing compel us to examine the issues
of competition and monopolies in the
nation’s payments system.

industry, it is also important to consider
the role of network effects in its operation.
To illustrate, consider the value of membership in a check clearinghouse to banks
with offices in a community where several
banks conduct business. Initially, each
bank sends payment instruments to each
of the other banks demanding payment in
currency for checks that are presented.
Then, two of the banks in the community
decide that they can reduce their operating
costs—and the risk of having currency lost
or stolen—by arranging for their messengers to meet at an intermediate point and
exchange checks. The two banks agree to
settle among themselves by debiting and
crediting balances they hold with each
other, rather than moving currency about
the community.
This clearing and settlement arrangement would be even more efficient if these
two banks were to get a third bank to join
them, clearing checks among the three
banks under rules they agree to adopt as a
clearinghouse. In the same way, this clearinghouse would be even more valuable for
its members if additional banks joined. If
these network effects are strong enough,
there will be one check clearinghouse in
the community, and all banks with offices
there will be members.
In an industry with network effects,
the first entity to develop a network has
an advantage over later entrants. To be
successful in developing a rival network,
the new entity must convince many participants to switch to its network simultaneously, since the value of a network to
each participant depends on the number
of other participants using the network.
Many years ago, the Reserve Banks developed a dominant network for interregional
check clearing, which gave the Federal
Reserve leverage over the operation of
the payments system. Even if some banks
did not like the rules under which the
Reserve Banks offered payment services
or the process innovations favored by the
Fed, those with a lot of interregional
checks to clear found it advantageous to
use the Fed’s clearing and settlement
network.

FEDERAL RESERVE BANK
SERVICES
This section reviews the laws and
Federal Reserve policies that govern the
activities of Reserve Banks as providers of
payment services. It also describes the
principal payment and payment-related
services provided by the Reserve Banks.
The appendix describes the payment
services of the Reserve Banks in more
detail and discusses major changes in the
services over the years.

FEDERAL RESERVE BANK

7

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Further, Reserve Banks have access to privileged supervisory information concerning
the condition of banks, which they may
use to protect themselves from losses in
providing payment services and related
credit. In addition, they can create reserves
to meet the liquidity needs of banks.

MCA GUIDELINES FOR PRICING
BANK SERVICES
The following section of the Monetary Control Act
(MCA) specified guidelines for the pricing of Reserve
Bank services:
“Over the long run, fees shall be established on the
basis of all direct and indirect costs actually incurred in
providing the Federal Reserve services priced, including interest on items credited prior to actual collection,
overhead, and an allocation of imputed costs which
takes into account the taxes that would have been paid
and the return on capital that would have been provided had the services been furnished by a private business firm, except that the pricing principles shall give
due regard to competitive factors and the provision of
an adequate level of such services nationwide.”
Thus, the basis for the Federal Reserve’s setting the
prices of its payment services below levels as specified
in this section of the MCA is inadequate competition in
markets for payment services or an inadequate level of
services in at least some regions of the nation.

Law and Policy Governing Reserve
Bank Services
The Reserve Banks provide payment
services under the authority of the Federal
Reserve Act, as amended over the years.
The terms and conditions under which
they provide services are governed by
regulations of the Board of Governors
and implemented through Reserve Bank
operating circulars.7
The Monetary Control Act of 1980
(MCA) was a watershed for the Reserve
Banks as providers of payment services.
Prior to passage of the MCA, the member
banks in the Federal Reserve shouldered a
required reserve burden which they could
satisfy through only two forms of nonearning assets: deposits held in accounts
with the Reserve Banks and vault cash.
Banks that were not members of the Fed
system were not burdened by this requirement. Provision of “free” payment services
by the Reserve Banks was viewed as an
offset to the reserve requirement burden.8
But the MCA changed all that:

Background

7

8

Operating circulars are detailed
instructions concerning particular banking services, including
account services, payment services, and the discount window.
Some apply uniformly to all 12
Reserve Banks, while others
apply to the services of individual Reserve Banks.
For an analysis of the value of
“free” Reserve Bank payment
services to member banks relative to the opportunity cost of
their required reserves, see
Gilbert (1977).

The Reserve Banks function as bankers’ banks: Banks that use the Fed’s payment services maintain reserve balances
at the Reserve Banks and have access to
credit from the Fed. All transactions cleared
through the Reserve Banks are settled on a
gross basis; that is, the value of each transaction is settled through a debit or credit
to a bank’s reserve account. Occasionally,
debits and credits resulting from use of the
Fed’s payment services can cause a bank
to miss its target for reserves, or they can
cause a negative reserve balance. Reserve
Banks lend reserves to banks that are temporarily short of funds, including both
intraday credit (daylight overdrafts) and
overnight credit (discount window loans
or overnight overdrafts). The Reserve
Banks also serve as fiscal agents to the federal government by providing payment
services to the United States Treasury and
to various other government agencies.
Reserve Banks have special privileges
and powers as suppliers of payment services.
For example, they have legal authority to
present checks for same-day settlement
later in the day than do private banks.

1. It extended the reserve requirements of the Federal Reserve to all
depository institutions.
2. It granted all depository institutions access to the discount window
and to Reserve Bank services.
3. It required the Reserve Banks to
charge explicit fees for their services.
Under the MCA, the Reserve Banks’
revenue from fees on their payment
services must, over the long run, equal or
exceed the cost of providing the services
plus a markup to reflect the tax rates and
profit rates of private-sector firms (see
shaded box on MCA guidelines). Thus,
the MCA subjects the Reserve Banks to

FEDERAL RESERVE BANK

8

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

market discipline similar to that faced by
commercial firms: Reserve Banks must
provide services efficiently, price them
competitively, and meet the market’s standards for quality services. Further, they
must be careful to gauge the profitability
of new service offerings.
Since passage of the MCA, the Federal
Reserve Board has also issued guidelines
that specify in more detail the conditions
under which Reserve Banks may provide
payment services. The guidelines issued in
June 1981, state that “the System should
be prepared to remove itself from the provision of those services that can be supplied more efficiently by the private sector,
unless there are overriding public interest
considerations for maintenance of an
operational presence by the System” (Federal Reserve Regulatory Service 7–191).
Further, the Board of Governors’ May 1990
policy statement on the role of the Federal
Reserve in the payments system sets additional conditions to be met before the
Reserve Banks may offer new payment services: “the service should be one that
other providers alone cannot be expected
to provide with reasonable effectiveness,
scope, and equity” (Federal Reserve Regulatory Service 7–145.1). Thus, the MCA,
together with Federal Reserve Board policies, establishes market-oriented criteria
for determining whether and how the
Reserve Banks are to provide services.
While these formal and explicit conditions under which Reserve Banks may
continue to offer existing services or enter
new payment markets were developed after
passage of the MCA, earlier decisions by the
Board of Governors suggest something
about the Federal Reserve’s philosophy in
providing services. In particular, there is evidence that, well before the MCA, the Board
wished to proscribe Reserve Bank involvement in the processing of new types of
payment instruments. In the second half of
the 1960s, for example, the Federal Reserve
came under some pressure from bankers to
adapt its check-clearing services to handle
the processing of credit card sales slips. For
a variety of reasons, including concern over
the public sector’s shouldering significant

new costs for handling a quasi-payment
instrument, the Board decided to deny the
credit card industry access to the check
clearing infrastructure of the Reserve Banks
(Brimmer, 1967). Accordingly, the Reserve
Banks play no role in processing credit card
transactions. Instead, a private-sector infrastructure has grown up to support this
important component of the payments
system.
In contrast to this decision on processing credit card slips, the Board agreed,
at approximately the same time, to requests
from bankers that the Reserve Banks provide operational support for the nascent
automated clearinghouse (ACH) as a
method of processing payments. The ACH
represented a desirable alternative to checks
that would require new automation systems and significant start-up costs. Because these start-up costs would have been
difficult for the private sector to absorb,
the Board permitted the Reserve Banks to
take on this new operational responsibility.

Payment Services of the
Reserve Banks
Cash Services. The Reserve Banks provide
coin and currency to banks on demand
and receive excess coin and currency from
banks; the banks’ reserve accounts are
debited and credited for the value of these
transactions. However, Reserve Banks do
not charge banks for cash services, since
the Board has determined that cash
services are a central bank function.
Check Clearing. The offices of Reserve Banks
throughout the nation receive checks from
banks for collection, and the proceeds from
the collection of these checks are credited to
the reserve accounts of the depositing banks.
The timing of credits reflects the length of
time required for the Reserve Banks to present
the checks to the banks on which they are
drawn (paying banks) and to receive payment, which is made by debiting the reserve
accounts of the paying banks. To facilitate this
process, the Reserve Banks operate a national
system for transporting checks to the paying
banks. Reserve Bank check-collection services
include both the clearing function (receiving

FEDERAL RESERVE BANK

9

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

CALCULATING THE FED’S SHARE
The calculations in Tables 1–3 are based on three categories of checks processed
by Federal Reserve Banks: those issued by the federal government, U.S. postal money
orders, and all other checks (referred to as “commercial checks”). Commercial checks
processed in the United States include “on-us” checks—checks drawn on the banks
where they were first deposited. Since “on-us” checks do not have to be cleared
between banks, in all three tables we have subtracted them from the totals for interbank checks processed in the United States.
For example, the figures for 1980 in Table 1 were calculated in this manner:
Annual Reports of the Board of Governors include the numbers of checks processed
by the Reserve Banks each year. The Annual Report for 1980 indicates that the
Reserve Banks processed 15,721 million commercial checks, 705 million government
checks, and 117 million postal money orders. The number for commercial checks,
however, reflects double counting: Checks received by one Reserve Bank that were
sent to another Reserve Bank for collection were counted as checks processed by each
bank. Beginning in 1982, the Annual Reports eliminated this double counting of commercial checks. The numbers for 1982 are available with and without the double
counting: The number of commercial checks processed without double counting is
94 percent of the number with double counting. Applying this 94 percent adjustment
to the data for 1980 yields an estimate of 14,777.7 million commercial checks, and
15,599.7 million for total checks, including government checks and postal money
orders.
The total number of commercial checks in the nation in 1980 is estimated as 42
percent of the number of commercial checks processed by the Reserve Banks. We
divided the number of checks the Fed processed by the percentage of checks that it
processed to arrive at the total number of commercial checks issued in the United
States in 1980:
14,777.7m/.42 = 35,185 million.
Of these 35,185 million checks, approximately 29.6 percent, or 10,414.8 million,
were “on-us” checks. We subtracted the “on-us” checks from the commercial checks,
then added the federal government checks and the U.S. Postal orders to arrive at the
total number of interbank payment items processed in 1980:
35,185m – 10,414.8m + 705m + 117m = 25,592.2 million.
In Table 2, the average estimated value of a check in 1980 was $792, and the average estimated value of an “on-us” check was $867. Average check values in these
tables are based on the values in the 1979 Atlanta Fed Check Study, adjusted for inflation and other factors. The values for checks processed by the Federal Reserve are
actual, except that 1980 data have been adjusted for the double counting that was
used in Federal Reserve reporting systems at that time. We multiplied the number of
commercial checks by the average value per check (35,185 million 3 $792) to arrive
at a total value of $27.9 trillion for checks processed in the United States in 1980. We
then subtracted the estimated value of “on-us” checks (10,414 million 3 $867 = $9.0
trillion) and added the value of federal government checks ($599 billion) and postal
money orders ($6 billion) to arrive at a total value of $19.4 trillion.
In Table 3, we divided the number of payment items that the Federal Reserve
Banks processed by the total number of interbank payment items processed in the
United States (as calculated in paragraph 3, above) to arrive at the Federal Reserve’s
share of payment-items processing for 1980:
15,599.7 million / 25,592.2 million = .61, or 61 percent.

FEDERAL RESERVE BANK

10

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

checks and presenting them to paying banks)
and the settlement function (debiting and
crediting reserve accounts).

government agencies. These services include issuing and redeeming U.S. Treasury
securities as well as securities of other U.S.
agencies.

Automated Clearinghouse. Banks that use
the Federal Reserve’s ACH service instruct
Reserve Banks to pay other banks (ACH
credit entries) or to receive payment from
other banks (ACH debit entries). These
entries are processed through the computer
facilities of the Reserve Banks (clearing function), and entries are posted to the reserve
accounts on the settlement dates designated
by the banks (settlement function).

Net Settlement. Banks that are members
of private clearing organizations may
decide to settle their mutual obligations
through multilateral netting. If a private
clearing service uses the net settlement
services of the Reserve Banks, the net debit
and credit positions of the private banks
are settled through entries to their reserve
accounts at the Reserve Banks.

Safekeeping of Definitive Securities and
Noncash Collection. Reserve Banks accept
definitive securities (securities in paper
form) for safekeeping. This service, however, is now largely limited to securities
used to collateralize government deposits
and discount window loans. The Reserve
Banks collect interest coupons and
matured securities and credit the proceeds
to the reserve accounts of banks that own
the securities.

TRENDS IN CLEARING AND
SETTLEMENT
The Reserve Banks’ share of total interbank payments has declined since 1980 for
at least three of the four principal types of
payment instruments: checks, large-value
funds transfers, and large-value securities
transfers. Tables 1, 2 and 3 show the declining Reserve Bank components for both
the volume and the value of interbank payment transactions. We have not been able
to develop a time series on the share of the
volume and value of ACH payments
processed by the Reserve Banks.

Wire Transfer of Funds. Banks with reserve accounts at Reserve Banks may initiate
transfers of their reserves to other banks
through the Fedwire funds transfer service.
Fedwire is a real-time gross settlement
system. Fedwire funds transfers are processed electronically and are final when
accepted for processing by the Reserve
Banks. A final payment is one which is
unconditional and irrevocable. Clearing
and settlement is virtually simultaneous.
Fedwire is described as a large-value funds
transfer service because it is designed to
facilitate interbank funds transfers (Horii
and Summers, 1994).

Check Clearing
Table 3 indicates a significant decline in
the Reserve Banks’ share of interbank check
clearing, in terms of both volume and
value.9 Between 1980 and 1994, the Reserve
Bank’s component of interbank checkclearing volume declined by about onethird, from an estimated 61.0 percent to
39.3 percent, while its check-value component declined from an estimated 48.5 percent to 24.9 percent. These declines are
consistent with a conventional interpretation of major changes in the interbank check
clearing market, including (1) the introduction of Reserve Bank pricing for services,
mandated by the MCA, (2) a fairly rapid
development of alternative private-sector
channels for check clearing, and (3) adoption by the Board of Governors of same-day
settlement amendments to Regulation CC.
These amendments to Regulation CC,
effective January 1, 1994, changed the

Wire Transfers of Securities. Ownership
of United States government securities and
some agency securities is recorded in the
securities accounts held by the Reserve Banks.
Banks can transfer ownership of these securities via the Fedwire securities transfer
service, and each transfer is final when accepted by the Reserve Banks for processing.
Fiscal Agency. The Reserve Banks provide
account, custodial, and payment services to
the U.S. Treasury and to a variety of other

FEDERAL RESERVE BANK

11

OF

S T. L O U I S

9

Interbank check clearings are
so-called “transit items,” for
which the payor (check writer)
and payee have accounts at different banks. These are in contrast to “on-us” checks, for
which the payor and payee
have accounts at the same
bank.

S E P T E M B E R /O C T O B E R 1996

Table 1

Volume of Interbank Non-Cash Transactions
(in millions of transactions)
1980

1985
Total

FR

1990
Total

FR

1994

Type of Payment

FR

Total

FR

Total

Check 1
ACH2
Large-Value
Funds Transfer 3
Securities Transfer 4
Card

15,599.7
227.0
25.8

25,592.2
––
39.0

16,687.0
585.0
45.0

33,489.8
––
69.9

19,304.0
1,435.0
62.6

39,670.0
––
99.9

17,149.0
2,379.0
72.0

43,637.4
2,521.8
117.5

––
N/A

––
––

7.7
N/A

7.7
––

10.9
N/A

12.9
10,478.1

12.6
N/A

19.1
13,681.0

Sources: Annual Reports of the Board of Governors of the Federal Reserve System and the Bank for International
Settlements.
1 See Shaded Box, page 10.
2 Total ACH volume represents Federal Reserve commercial and government items plus items processed exclusively by private-sector
arrangements. The figures for Federal Reserve volumes are taken from actual, recorded data. The source of the estimate for the private-sector volume is the National Automated Clearing House Association. The private ACH processors active in 1994 included the
Arizona ACH, Hawaii ACH, New York ACH, and Visa ACH. Data for private ACH processors for periods before 1994 are either not available or incomplete. Note that the majority of items handled by private ACH processors are also delivered to the Federal Reserve for
processing, to gain access to endpoints serviced only by the Federal Reserve. In 1994, for example, the total number of items actually
originated and received by private ACH operators was estimated to be 521 million; of these, only 143 million were also delivered
exclusively within the private arrangements.
3 The total volume of large-value funds transfers is the sum of Fedwire funds transfers and Clearing House Interbank Payments System
(CHIPS) transfers.
4 Total number of Fedwire securities transfers plus adjusted gross volume estimates for the securities transfers of the Government
Securities Clearing Corporation (GSCC) and the Participants Trust Company (PTC). The GSCC estimates were adjusted downward by subtracting the number of end-of-cycle transfers made through Fedwire, to avoid double counting. All securities transfers of the
Government National Mortgage Association (GNMA) were processed through PTC; they could, however, have been processed by Fedwire,
had the Federal Reserve chosen to provide such services to GNMA.

return the checks or pay the collecting
bank through a Fedwire funds transfer by
the close of business the same day. The
paying bank is not allowed to charge the
collecting bank a fee for same-day settlement. Banks may waive these rules for the
timing of check presentment and means
of payment if they wish (Fitzgerald and
Macoy, 1993; Crockett, 1994b). Reserve
Bank check collection volume through
September 1994 was 12 percent below the
volume for the same period in 1993. This
decline is attributed largely to same-day
settlement (Marjanovic, 1994b).

rules under which banks pay each other
for checks. Before implementation of the
same-day settlement provisions, a bank
presented with checks directly by another
private bank could either pay the collecting bank the following business day or
charge the bank a fee for payment the
same day. Reserve Banks, in contrast, debited the reserve accounts of paying banks
the same day they delivered the checks to
the banks, and the Reserve Banks did not
pay fees for this privilege. When private
correspondents complained that these
rules gave the Reserve Banks an unfair
advantage, the Fed adopted the same-day
settlement regulation, which says that if a
collecting bank presents checks to the
place of business of a paying bank before
8 a.m. local time, the paying bank must

Automated Clearing House (ACH)
The Reserve Banks are the dominant
processors of ACH payments. They handle
all government-related transactions and a

FEDERAL RESERVE BANK

12

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Table 2

Value of Interbank Non-Cash Transactions
(in trillions of dollars)
1980

1985

1990

1994

Type of Payment

FR

Total

FR

Total

FR

Total

FR

Total

Check1
ACH2
Large-Value
Funds Transfer3
Securities Transfer4
Card

9.4
0.3
47.9

19.4
––
85.0

10.1
2.1
109.1

31.9
––
187.5

13.2
4.7
199.1

43.5
––
421.1

12.6
8.4
211.2

50.6
9.1
506.6

––
N/A

––
0.1

74.2
N/A

74.5
0.2

99.9
N/A

108.1
0.5

144.7
N/A

170.0
0.7

Sources: Annual Reports of the Board of Governors of the Federal Reserve System and the Bank for International
Settlements.
1 See Shaded Box, p. 10.
2 The value of transactions handled by the Federal Reserve plus the value of transactions handled solely by private ACH processors (see
Table 1, footnote 2). For 1994, the estimated value of ACH transactions processed solely by the private sector was about $700 billion.
3 The sum of the value of Fedwire funds transfers and Clearing House Interbank Payments System (CHIPS) transfers.
4 The sum of the value of Fedwire securities transfers, plus the value of the adjusted gross volume for the Government Securities Clearing
Corporation (GSCC) plus the value of Participants Trust Company (PTC) adjusted gross volume. See Table 1, footnote 4, for more details.

Table 3

Federal Reserve Share of Interbank Non-Cash Transactions
Check
Year
1980
1985
1990
1994

ACH

Volume

Value

Volume

Value

(percent)

(percent)

(percent)

(percent)

61.0
49.8
48.7
39.3

48.5
31.7
30.3
24.9

––
––
––
94.3

Large-Value Funds Transfers
Volume
Value

––
––
––
92.3

large share of commercial transactions.
However, their present share of the volume
and value of interbank ACH transactions,
which exceeds 90 percent, does not appear
to be sustainable. In addition to the Reserve Banks, three private organizations
process ACH payments: the Arizona
Clearing House Exchange, the New York
Clearing House, and Visa USA. The financial press cites the share of ACH payments
processed by the Reserve Banks at about
80 percent, with these three organizations
processing the remaining 20 percent (Marjanovic, 1995a,b). This statistic is based on

Securities Transfers
Volume
Value

(percent)

(percent)

(percent)

(percent)

66.2
64.4
62.7
61.3

56.4
58.2
47.3
41.7

––
100
84.5
66.0

––
100
92.4
85.1

the fact that these private organizations
receive 20 percent of total ACH entries.
However, some of the ACH entries these private organizations receive are routed through
the Reserve Banks for processing.
In this paper, the volume of ACH payments attributed to the Reserve Banks is
that actually processed by the Reserve
Banks, whether the originating institutions
delivered the information on ACH entries
to the Reserve Banks or to private processors. This method of calculating the component of ACH payments processed by
the Reserve Banks is consistent with the

FEDERAL RESERVE BANK

13

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

method of calculating the component of
checks processed by the Reserve Banks. For
instance, checks counted as processed by
the Reserve Banks include those deposited
by the banks of first deposit, and by banks
that serve as intermediary correspondents
for the banks of first deposit.
The high Reserve Bank share can be
attributed to the unique circumstances
surrounding the development of the ACH
payment mechanism, which was initially
subsidized by the Reserve Banks. Significant developments in the market for ACH
services in the last five years, relating to
changes in technology, banking structure
and the entry of private providers, will
almost surely combine to reduce the proportion of ACH payments processed by
the Reserve Banks.

1994. The volume component fell from
66.2 percent to 61.3 percent, and the value
component fell from 56.4 percent to 41.7
percent. While the reasons for the rise in
the CHIPS component of large-value funds
transfers are complex, they are related in
part to the rapid growth of international
payments. The Reserve Banks have not
considered the settlement of foreign
exchange and other international transactions to be part of the mission of Fedwire,
and, therefore, they have not attempted to
design the Fedwire service to meet the
specific funds-transfer needs of that
part of the market. They have, however,
responded to the market for fundstransfer services, and to new recordkeeping requirements resulting from
anti-money-laundering legislation, by
adopting a new format for funds transfers
over Fedwire that is based on the standards of the Society of Worldwide Interbank Finance Telecommunications (SWIFT).
Conversion to the new format will be
completed by the end of 1997.
Another factor that may have reduced
the Reserve Banks’ share of large-value
funds transfers is the Federal Reserve’s payments system risk-reduction program,
which in recent years has increased the
appeal of multilateral netting for banks. The
risk-reduction program, which has placed
significant emphasis on containing the
amount of intraday credit provided by the
Reserve Banks, has probably stimulated use
of alternatives to Fedwire for clearing largevalue transactions.10 On the other hand, the
risk controls adopted by CHIPS, which have
increased the cost of funds transfers over
that system, have tended to offset the effects
of the Reserve Bank’s risk-reduction
measures.

Large-Value Funds Transfer

10

For a description of the policy
of the Federal Reserve on daylight overdrafts and payments
system risk, see Richards
(1995).

The Reserve Banks guarantee finality
of funds transfers among banks over Fedwire; private banks that receive funds
transfers over Fedwire do not have to be
concerned that the transfers will be reversed by the Reserve Banks because of the
failure of the sending banks to fund their
payments through the Reserve Banks.
Casual observers of the market for largevalue funds transfer might conclude that
the Federal Reserve would have a virtual
monopoly on this service. The information
in Tables 1, 2, and 3, however, indicates
that this conclusion would be incorrect.
Large banks that are members of the Clearing House Interbank Payments System
(CHIPS)—a wholesale wire-transfer network owned and operated by the New
York Clearing House—use that system as
an alternative to transfers over Fedwire for
large-value funds transfers. Members of
CHIPS net their interbank obligations
multilaterally and settle these obligations
as a group at the end of the day using Fedwire funds transfers. They use CHIPS
largely for settling the dollar side of foreign exchange and for other international
transactions.
The component of total large-value
funds transfers (over Fedwire and CHIPS)
handled by Fedwire declined from 1980 to

Securities Transfer
As with funds transfer, the casual
observer might conclude that the Federal
Reserve has a virtual lock on the market
for securities transfers. In fact, however,
the Federal Reserve has restricted the
range of U.S. Treasury securities and
agency securities for which the Reserve
Banks serve as depositories and provide

FEDERAL RESERVE BANK

14

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

transfer services. As a result, private systems for clearing and settling transactions
involving these securities have developed.
For example, the Participants Trust Company (PTC) now serves the entire market
for clearing and settlement of Government
National Mortgage Association (GNMA)
securities.
Private-sector arrangements for netting securities transactions are also becoming more attractive to banks. The Government Securities Clearing Corporation
(GSCC) has developed a multilateral netting service for future-dated U.S. government securities transactions, and it is in
the process of testing an enhanced service
that will support netting for same-day
transactions. Introduction of this service
by GSCC will likely trigger a significant
further decline in the Reserve Banks’ share
of government and agency securities transfers. Table 3 indicates that Fedwire’s share
of the volume of securities transfers declined from 100 percent to 66.0 percent
between 1985 and 1994. The Reserve Bank
share of the value of securities transfers fell
from 100 percent to 85.1 percent over the
same period.

ernment securities transfers, can be
expected to cause significant further
reductions in the Reserve Bank component
of large-value securities transfers. In addition, continued increases in the use of new
types of retail (small-value) payments
instruments, in which the Reserve Banks
are not active, could erode further their
role in the processing of retail payments
generally.

REASONS FOR THE
RESERVE BANKS’
DECLINING SHARE OF
INTERBANK PAYMENTS
Tables 1, 2 and 3 indicate that the
Reserve Banks’ historically important role
in providing clearing services has been
declining, although it is still significant.
Changes in technology and in banking
structure have reduced the Reserve Banks’
advantages in providing the dominant network for clearing and settlement of payments. In addition, the policies of the
Board of Governors have stimulated a
greater role for the private sector in
clearing interbank payments.

Technology

Card Transactions

New technology is perhaps the single
most important force leading to new initiatives for processing payments in the
private sector. Within the last decade or
so, the costs of both computer processing
and data communications have fallen
dramatically. As a result, automated processing systems are now within the financial reach of individual institutions as well
as private clearinghouses. At one time,
ACH processing required large mainframe
computer systems. Now, very powerful,
small, and relatively inexpensive microprocessors are able to handle large
volumes of transactions. Moreover, valueadded networks offer a wealth of national
and even international data communications pathways, including networks with
sufficient control and security features to
handle electronic payment transactions.
Thus, dramatic reductions in costs have
facilitated the development of alternative
networks for payment processing func-

As we noted earlier, the Reserve Banks
do not clear payments made by cards.
Table 1 shows that credit-card transactions
have grown rapidly in recent years and by
1994 accounted for about 18 percent of
the number of payments made by credit
card, check and ACH. If payments based
on other types of cards, such as debit
cards and stored-value cards, grow rapidly relative to older types of payments
instruments, the percentage of all retail
payments processed by the Reserve
Banks can be expected to continue
declining.
In summary, the Reserve Banks’
components of both the volume and the
value of interbank payments have declined
for small-value retail and large-value
wholesale funds and securities transactions
since about 1980. Prospective developments, including introduction by GSCC
of multilateral netting for same-day gov-

FEDERAL RESERVE BANK

15

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

change large volumes of checks and other
payment instruments directly, without
relying on a processing intermediary.

tions that were once the primary domain
of the Reserve Banks.

Banking Structure

11

Berger and Humphrey (1988)
conclude that nationwide interstate banking would reduce the
resources used in check collection. In addition, they estimate
that it would reduce the share
of total checks processed by
Reserve Banks by between 43
and 60 percent over a 10-year
period.

Federal Reserve Policy

Regional interstate banking has also
reduced the advantages of the Federal
Reserve in interregional check clearing.
Banking concentration resulting from the
rise in interstate banking has increased the
proportion of transactions handled by
banks as “on-us” transactions, which bypass interbank clearing and settlement
channels. Further, bank holding companies
have been able to organize payments clearing among their affiliated banks on a regional basis, often by establishing regional
processing centers. Today, the country is
experiencing a major new interstate banking movement as a result of the Riegle-Neal
Interstate Banking and Branching Efficiency
Act of 1994. More and more, interstate
banking extends network efficiencies to private institutions throughout the country,
thus eroding the Fed’s interregional checkclearing advantage.11
The last 10 or 15 years have seen the
formation of national clearinghouses for
both paper and electronic transactions. For
example, in 1991, Visa USA began offering a
national ACH processing service. Further,
the New York Clearing House has expressed
interest in expanding the geographic scope
of its ACH service and connecting its processing network to other private sector
providers, such as Visa (Marjanovic,
1995a,b). With respect to check clearing,
the National Clearing House Association,
formed in 1992, arranged the clearing of an
estimated 2 million checks per day in 1994
(Marjanovic, 1994a). Similarly, the Electronic Check Clearing House Organization
(ECCHO) which was introduced in 1990,
was clearing an estimated 1 million checks
per day by 1994, on the basis of electronic
cash letters (Crockett, 1994a). Banks are
also active in establishing consortia to
exchange retail transactions in ATM and
point-of-sale networks. Formation of these
private networks for clearing payments
reflects, to some extent, the increased concentration of banking in recent years. A few
large banking organizations now can ex-

The Federal Reserve has strongly
embraced market techniques that promote
more efficient payment operations. Explicit pricing of Federal Reserve payment
services, introduced by the MCA, has eliminated subsidies to banks that use Reserve
Bank services. In addition, explicit pricing
of intraday, reserve-account overdrafts has
increased the costs of using Reserve Bank
services for banks with relatively large
intra-day overdrafts (Richards, 1995).
Finally, the Federal Reserve Board’s action
in January 1994 requiring same-day check
settlement helped to reduce the barriers to
check clearing between private parties.
Accordingly, within the last 15 years, the
explicit cost of using Federal Reserve payment-processing services has been put on
more comparable terms with private-sector
alternatives, and artificial legal barriers to
private clearing have been removed.
Federal Reserve Board policy has also
limited the involvement of the Reserve
Banks in the payments system by restricting the scope of their services. For instance, the Federal Reserve declined to provide services for clearing and settling credit
card sales slips and book-entry transfers of
GNMA securities. The limitations on the
scope of Reserve Bank payment services
have facilitated the development of private
systems for clearing and settling payments.

Outlook for Fed Payment Services
Technology, banking structure, and
Federal Reserve policy will likely continue
to influence payments processing, certainly for the foreseeable future. The outcome
is likely to be a continuation of the trends
shown in Table 3—declines in the components of various types of interbank
payments processed by the Reserve Banks.
Indeed, because virtually all the factors
discussed in this paper have emerged relatively recently, the trends in Table 3 could
accelerate, at least for small-value payments such as checks and ACH.

FEDERAL RESERVE BANK

16

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

SOME IMPLICATIONS OF
REDUCTIONS IN RESERVE
BANK SERVICES

however, might also lead to distortions in
the pricing of payment services that
would result from collusive behavior by
a few large, nationwide branching organizations.

Trends in the processing of payments
by Reserve Banks have implications for the
efficiency of the payments system and for
the risk of disruptions in the operation of
the payments system.

Federal Reserve as Payments System Rule
Maker. The success of innovations in
improving the efficiency of the payments
system requires cooperation among
providers of payment services. For instance, the Fed and the banking industry
agreed many years ago to encode checks
with magnetic characters (the MICR line)
that make it possible for banks to sort
checks by machine. This innovation would
have been of little value if it had been
adopted by only a few banks. Another example involves the truncation of checks in
the collection process: To maximize the
benefits of truncation (in which the actual
paper check is taken out of circulation),
the first bank that handles a check would
convert the paper instrument into an electronic instrument and send the payment
information on the check through the collection system electronically. Such an innovation would require the cooperation of
virtually all banks.
Prior to 1914, the payments system of
the United States functioned without a
mechanism by which banks could cooperate in adopting innovations to make
interregional check collection more efficient. Clearinghouses played such a role in
their local communities. There was, however, no national clearinghouse to coordinate change for the national payments
system. Banks collected checks drawn on
banks located in distant cities through a
correspondent banking system that often
routed checks to paying banks indirectly,
to avoid the exchange charges of paying
banks. Indirect routing of checks increased
the expense and length of time in check
collection. Since its formation, the
Federal Reserve has functioned as the
de facto national coordinator of the
payments system. The role of Reserve
Banks as major providers of payment services has been important in facilitating a
number of improvements in the efficiency

Implications for Efficiency
Subsidy for use of Reserve Bank Services.
Prior to 1980, the Reserve Banks did not
charge member banks explicit fees for use
of their payment services. In the 1970s,
the Reserve Banks may have compounded
the inefficient use of resources in the payments system by subsidizing the collection
of local checks through the establishment
of Regional Check Processing Centers
(RCPCs). Because RCPCs provided sameday check crediting to the reserve accounts
of collecting banks for checks drawn upon
banks located in the same area, many local
check clearinghouses could not compete
and closed down (Frodin, 1984). While
the establishment of RCPCs may have accelerated the speed of collection, it created
additional incentives for banks to use the
Fed for check clearing rather than continuing direct exchanges of checks among
nearby banks.12 Recent declines in the use of
Fed payment services reflect more efficient
use of resources resulting from the elimination of the subsidy provided by “free”
Reserve Bank services.
Implications of Interstate Banking. Inefficiency in the check-collection system prior
to 1914 reflected, to a large extent, the lack
of nationwide banking organizations. High
exchange charges and lengthy delays in
check collection that resulted from arrangements to avoid exchange charges would
have been reduced or eliminated by nationwide banking.13 The spread of nationwide
interstate banking reduces the chances that
the declining role of the Reserve Banks in
payment processing will produce a return
to the kind of payments system inefficiency that existed prior to the formation
of the Federal Reserve. Interstate banking,

FEDERAL RESERVE BANK

17

OF

S T. L O U I S

12

For a description of RCPCs and
analysis of their implications for
the efficiency of the payments
system, see Morris (1974,
1975a,b); White and
Torgerson (1974); and
Viswanathan and Mayo
(1975).

13

Jessup (1967) reports that
nonpar banks (those imposing
exchange charges) tended to
be located in unit banking
states. This observation supports the claim that nationwide
branch banking would have
reduced or eliminated exchange
charges.

S E P T E M B E R /O C T O B E R 1996

tional role for the Reserve Banks, the
Federal Reserve Board, with its broad regulatory power, can continue to promote
payments system efficiency. An important
example of the Board’s use of this regulatory authority to promote efficiency in
the check-collection system is its recent
introduction of same-day settlement for
checks, which took effect in January 1994.
The Fed’s same-day settlement regulation
sets the rules under which collecting
banks present checks and receive payment
the same day without paying fees to paying
banks. The authority for this action of the
Board is derived from the Expedited Funds
Availability Act of 1987, which granted the
Board regulatory authority over interbank
payment relationships for purposes of promoting efficiency of the payments system.
Implementation of the change did not require a large operational role for the Reserve Banks; in fact, it has caused a decline
in the check collection volumes of the
Reserve Banks (Marjanovic, 1994b).
Increased concentration of the banking industry through interstate banking
can facilitate innovation through cooperation among the banks themselves, independent of the Fed’s efforts. The evidence
suggests that, in the past, there were too
many banks for effective cooperation.
Associations of relatively small numbers of
large banks, however, can work out agreements on innovations that benefit a majority of their members. For instance, the
banks that formed ECCHO agreed to accept electronic transmission of information
about checks as legal presentment. In
March 1995, the New York Clearing House
announced that its members had reached
similar agreement (Marjanovic, 1995c).
These examples illustrate innovation in
the payments system through voluntary
association.
Large numbers of banks are able to
coordinate the clearing and settlement of
payment instruments other than checks,
and to adopt innovations. For example,
Visa and Mastercard coordinate their payment operations for thousands of their
member banks. The growth of regional
ATM networks indicates that many

SAME-DAY SETTLEMENT: RISK
REDUCTION FOR COLLECTING
BANKS
Prior to the implementation of the Federal
Reserve’s same-day settlement regulation, banks that
collected checks through correspondents generally
received payment in the form of credits to their balances at correspondent banks. Under the new regulation respondent banks can send checks directly to
paying banks and receive payment the same day via
wire transfers of funds to their reserve accounts at
Reserve Banks. Adoption of same-day settlement, therefore, gives respondent banks more options for limiting
their exposure during periods of financial distress. For
relatively small banks, however, the cost of collecting
checks through direct presentment may exceed the risk
of collecting through correspondents. The operating
costs of collecting checks through direct presentment
include the costs of sorting checks and arranging for
couriers to present them directly to the paying banks.
Even in times of financial stress in the banking industry, relatively small banks that rely on correspondents
for check collection are more likely to continue using
correspondent bank services than to switch their checkcollection operations to direct presentment.
of the payments system, including the following:
1. Elimination of nonpar banking.14
14

Some authors challenge the
idea that actions of the Reserve
Banks to eliminate nonpar
banking improved the operation
of the payments system. See
Baxter (1983). The literature
on exhange charges provides
conflicting views of this subject.
See Frankel (1995); Gilbert
(1991); and Salop (1990).
For purposes of this section, it
is sufficient to argue that, given
the limited power Congress
granted to the Reserve Banks
over the operation of the payments system, the Federal
Reserve was effective in establishing clearance of checks at
par as the standard for the
banking industry only because
of the major role of the Reserve
Banks in check clearing and
settlement.

2. Addition of the MICR line to
checks, making them readable by
check-sorting machines.
3. Creation of the automated clearinghouse.
4. Expedited processing of return
items after passage of the Expedited
Funds Availability Act.
5. General promotion of check imaging and electronic presentment
(Marjanovic, 1996).
Will the benefits of this leadership and
innovation be lost if the Reserve Banks
have a substantially smaller role in the
processing of payments in the future?
Not necessarily. Even with a smaller opera-

FEDERAL RESERVE BANK

18

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

bankers can work together to provide ATM
services for their customers. The National
Automated Clearinghouse Association
(NACHA) sets rules and standards for
ACH, even though it does not itself process ACH payment items. Finally, there are
well established institutional arrangements
for setting standards for various aspects of
the payments system. Thus, past experience with check collection in the United
States may exaggerate somewhat the
importance of the Reserve Banks as providers of payment services in facilitating
innovation in the payments system.

If the Reserve Banks’ check collection
volumes fall substantially in the future,
and their capacity to clear checks is reduced accordingly, they may no longer be
able to fulfill the role of check clearer of
last resort. In periods of financial stress,
this situation could put an extra burden of
responsibility on the banking industry for
ensuring the safe operation of the payments
system. Banks collecting checks would
need to be vigilant in managing their risk
when choosing correspondents and in
agreeing to forms of settlement for checks
presented directly to paying institutions.
In this context, the Federal Reserve’s
Regulation F, “Limitations on Interbank
Liabilities,” mandates careful management
of such interbank relationships.

Access to Payments Systems. When the
Reserve Banks provided the dominant
nationwide system for banks to clear and
settle payments, access to the payments
system was determined by legislation and
the operating rules of the Reserve Banks.
As private organizations emerge to rival
the Reserve Bank’s nationwide clearing and
settlement arrangement, access will be
determined at least in part by these private
organizations. Various agencies of the government and the courts might become
involved in settling disputes on the conditions under which private arrangements
for clearing and settling payments may
exclude some providers of payment
services.15

Implications for Risk
Does the declining role of the Reserve
Banks in processing payments increase the
risk of payments system disruption? The
answer depends on the nature of the
shocks to the payments system.
Bank Runs. Prior to the formation of the
Federal Reserve System, depositor runs
were the most important source of risk to
banks. The Fed can deal with threats originating from depositor runs by injecting
reserves into the banking system through
open-market operations and discount
window loans. In addition, federal deposit
insurance limits the vulnerability of banks
to depositor runs.

Check Clearer of Last Resort. Issues
raised by the role of the Reserve Banks as
check clearer of last resort have implications for both efficiency and risk. The rise
in Reserve Bank check clearing during the
Texas banking crisis during the second
half of the 1980s and early 1990s illustrates the role of Reserve Banks as check
clearer of last resort (Clair, Kolson and
Robinson, 1995). When major banks headquartered in the Southwest were in serious
financial trouble, respondents turned to
the Fed for check collection because they
did not want to suffer disruptions and possible losses resulting from the failure of
their correspondents (see shaded box,
page 18). Given their major role in check
processing, the Reserve Banks can absorb
additional check volume when circumstances disrupt other check-collection channels.

Securities Transfers. Another possible
shock to the payments system would be
the disruption of arrangements for transferring ownership of securities. Is it
important that the Reserve Banks retain
a major role in processing securities
transfers in order to minimize the effects
of such shocks on the payments system?
Alternatively, are private arrangements
for securities transfers sufficiently
sound to minimize the chances of such
shocks?
Parties to securities transactions must
be able to trust their agents to perform
as contracted. For instance, individual
investors in corporate stock must trust

FEDERAL RESERVE BANK

19

OF

S T. L O U I S

15

See Carlton and Frankel
(1995) for analysis of a court
case involving a dispute over
access of a bank to Visa for
issuing credit cards. Carlton and
Salop (1996) discuss the issue
of access by firms to joint ventures in a variety of cases.

S E P T E M B E R /O C T O B E R 1996

their brokers to execute trades according
to their orders. They must also trust that
the organizations established to clear
trades and settle obligations among brokers will be effective in settling trades.
For investors in U.S. Treasury and agency
securities, banks function as their agents
by holding securities with the Reserve
Banks and with private depositories. When
the investors decide to sell, the banks use
Fedwire or a private system to transfer
ownership of the securities and settle the
trades. For investors in securities transferred through privately operated systems,
the risk of not receiving the securities
they have paid for, and the risk of not
receiving cash for securities they have
sold, depends on the reliability of netting
arrangements among members of the
systems.
Securities transfers through private
systems have not created problems for the
operation of financial markets, because
these systems are well designed. In addition, active oversight by authorities such
as the Federal Reserve has ensured that
such private arrangements have the controls and guarantees needed to make them
reliable. In particular, the guidelines for
operation of delivery vs. payment systems
that were released by the Federal Reserve
Board on June 15, 1989, refer to various
controls, including liquidity safeguards,
credit safeguards, and open-settlement
accounting.
As long as the private systems for
securities transfers are appropriately supervised and maintain adequate risk controls,
the migration of securities transfers from
the Reserve Banks to private systems
would not appear to create problems for
the operation of financial markets. The following sections indicate, however, why the
Fedwire service for transferring funds and
securities remains essential for the settlement of obligations among members of
these private systems. These sections also
discuss the limited authority of the Federal
Reserve over the operations of private
clearing organizations which is derived
from the role of the Reserve Banks in providing settlement services.

Settlement Using Liabilities of Private
Banks: Moral Hazard and Systemic Risk.
Another source of shock to the U.S. payments system could result from the failure
of a major bank used for settlement by a
significant number of other banks. Prior
to 1914, banks settled payment obligations among themselves by transferring
ownership of deposit liabilities at private
banks, and major disruptions occurred
when customers lost confidence in the
nation’s money center banks. Given the
declining role of the Reserve Banks in processing payments, the future might bring
increased public reliance on a few large
banks for settling payment obligations. If
it does, the government might need to
ensure the survival of those banks, to prevent disruption of the payments system.
This reliance on a few large banks at the
heart of the payments system could
amplify any moral hazard in bank supervision and regulation. Since the failure of
the bank would be too disruptive to the
payments system, participants in the
financial system could assume that there
would be little risk in transactions with
those banks, including the purchase of
their short-term liabilities.
Multilateral Clearing Arrangements and
Systemic Risk. Systems for clearing payments among banks can be designed to
avoid the moral hazard outlined above.
Consider, for instance, the design of
CHIPS. Federal Reserve policies that apply
to the operation of private large-dollar
funds-transfer systems such as CHIPS
include the requirement that such systems
have means to ensure settlement in the
event of a default by a major participant.
Since these arrangements would prevent
major disruptions in the payments system
in the event of the failure of a particular
bank, investors in bank equities and liabilities cannot assume that any one bank is
essential to the operation of the payments
system.
We argue that the Reserve Banks must
continue to offer Fedwire services to facilitate access to reserve accounts and especially, to ensure the integrity of net settle-

FEDERAL RESERVE BANK

20

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

ment arrangements. Banks that use the net
settlement services of Reserve Banks in
their transactions with private paymentsclearing systems must have a mechanism
like Fedwire for transferring reserves, to
cover their net debit positions. At times,
these banks must borrow reserves from
each other to cover their net debits. Fedwire enables them to transfer securities
electronically, both to provide collateral to
lenders, and to post collateral with Reserve
Banks for discount window borrowing.
Authority of the Federal Reserve over
the operations of private clearing organizations rests principally on the role of the
Reserve Banks as providers of settlement
services, since the Fed has no statutory
authority for central bank oversight of private clearing organizations. As a service
provider, the Federal Reserve can make
safe-and-sound operation of clearing organizations a condition to their using its
interbank settlement services. The Board
stated its standards for the operation of
private clearing organizations in December 1994, in a policy statement titled
“Privately-Operated Large-Dollar Multilateral Netting Systems.” The threat of
discontinuing its settlement support for
such clearing organizations, however, is a
very blunt supervisory instrument. For
example, the Fed could disrupt clearinghouse operations, and therefore the
payments system, by withdrawing its
settlement services. The simple fact that
the Fed could itself trigger an immediate
operational crisis by withdrawing support
for settlement calls into question the
Federal Reserve’s willingness ever to
invoke such a harsh action.
The trends in the U.S. dollar payments
system described in this paper indicate a
major shift toward greater reliance on private arrangements for clearing both smalldollar and large-dollar payments. Increased
privatization of the U.S. dollar payments
system and a concomitant decline in the
operational role of the Federal Reserve
Banks raise questions about the adequacy
of the Federal Reserve’s supervisory
authority to fulfill the original Congressional mandate for ensuring the stability of

the nation’s payments system. In fact, the
Federal Reserve appears to be somewhat
unusual among central banks in that it
does not have explicit statutory powers
related to the supervision of clearing organizations. Some of the private clearing
organizations have implemented new risk
controls to ensure settlement in the event
of default by any of their members. These
actions indicate some of the Fed’s regulatory clout under current limitations on its
statutory authority. It is unclear, however,
whether the Fed’s indirect influence on
private clearing organizations through its
role as provider of settlement services will
be sufficient to ensure the safety and soundness of the payments system in the future.

CONCLUSIONS
The Federal Reserve Banks’ role in
processing payments—in terms of both
volume and value—has declined since
1980, when Congress enacted legislation
requiring the Reserve Banks to charge for
their payment services. This decline can be
expected to continue or even accelerate in
the future. While the declines in the shares
of payments processed by the Reserve Banks
following pricing of the services represent a
more efficient use of payments system resources, the declining role of the Federal
Reserve Banks in payments processing has
other important implications for the efficiency and stability of the payments system.
One of these implications relates to
innovation. In the past, the actions of the
Reserve Banks to foster innovation in the
payments system relied on the status of the
Reserve Banks as major providers of payment services. Will the Reserve Banks’
declining role in payments processing
eliminate the Fed’s leadership in innovation? Not necessarily. The Federal Reserve
Board has broad authority to promote safe
and efficient payment methods undertaken
bilaterally between depository institutions,
especially in the check-collection system.
This authority is independent of the Reserve Banks’ operating role in the payments system. In addition, the growing
concentration of the banking industry
through interstate banking is facilitating

FEDERAL RESERVE BANK

21

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

innovation by cooperating groups of banks
that would have been more difficult when
there were many more separate banking
organizations.
At the same time, the growth of private payment networks raises some important new issues with respect to competition. In the past, access to the payments
system was determined largely by law and
by Federal Reserve policies; now some of
the issues involving access to the new private payments arrangements will be settled
in the courts.
Does the declining role of Reserve
Banks in processing payments increase the
risk of disruption in the operation of the
payments system? The answer depends on
how banks in bilateral and multilateral
private clearing arrangements settle their
obligations. Settlement through debits and
credits to accounts at private banks would
make the system vulnerable to disruption
in the event of sudden failure by banks
that provide settlement services. Fortunately,
the settlement services of the Reserve Banks
can limit this risk—to banks, and to the
Federal Reserve in its role as lender of last
resort. To facilitate the use of reserves for
interbank settlement, whether net or gross,
Reserve Banks should continue offering
Fedwire funds and securities transfer services and net settlement services. The
Federal Reserve System is able to influence
the practices of clearinghouses primarily
by setting conditions for their use of the
settlement services of the Reserve Banks.
The Fed does not have statutory authority
to act as the supervisor of clearing organizations. It is not clear at this time whether
the Fed’s limited influence over clearing
organizations will be adequate to maintain
the safety and soundness of the payments
system as the share of payments cleared
through private channels continues to rise.

Baxter, William F. “Bank Interchange of Transactional Paper: Legal and
Economic Perspectives,” Journal of Law and Economics (October
1983), pp. 541-88.
Berger, Allen N., and David B. Humphrey. “Interstate Banking and the
Payments System,” Journal of Financial Services Research (1988),
pp. 131-45.
Blommestein, Hans J., and Bruce J. Summers. “Banking and the
Payment System,” in The Payment System: Design, Management,
and Supervision, Bruce J. Summers, ed. International Monetary Fund,
1994, pp. 15–28.
Brimmer, Andrew F. “Bank Credit Cards and Check-Credit Plans:
Development and Implications.” Remarks presented to a Joint
Luncheon of Commercial Bankers and the Board of Directors of the
Federal Reserve Bank of San Francisco, August 3, 1967.
Carlton, Dennis W., and Alan S. Frankel. “Antitrust and Payment
Technologies,” this Review (November/December 1995), pp. 41-54.
Carlton, Dennis W., and Steven C. Salop. “You Keep on Knocking but
You Can’t Come In: Evaluating Restrictions on Access to Input Joint
Ventures,” Harvard Journal of Law and Technology (Summer 1996),
pp. 1-34.
Clair, Robert T., Joanna O. Kolson, and Kenneth J. Robinson. “The Texas
Banking Crisis and the Payments System,” Economic Review, Federal
Reserve Bank of Dallas (First Quarter 1995), pp. 13-21.
Crockett, Barton. “Electronic Bad-Check Notices for Corporations
Expected Soon,” The American Banker (January 6, 1994a), p. 17.
______. “Fast Clearing Off to Slow Start, Poll Finds,” The American
Banker (May 12, 1994b), p. 18.
Dewald, William G. “The National Monetary Commission: A Look
Back,” Journal of Money, Credit and Banking (November 1972), pp.
930-56.
Dwyer, Gerald P., Jr., and R. Alton Gilbert. “Bank Runs and Private
Remedies,” this Review (May/June 1989), pp. 43-61.
Economides, Nicholas, and Lawrence J. White. “Networks and
Compatibility: Implications for Antitrust,” European Economic Review
(1994), pp. 651-62.
“Federal Reserve in the Payments System,” 7-145.1. Federal Reserve
Regulatory Service, 7•48.
“Federal Reserve System Guidelines for the Provision of Financial
Services,” 7-191. Federal Reserve Regulatory Service, 7•51.
Fitzergerald, Robert M., and Ian W. Macoy. “Check-Processing Rule
Means Efficiency, Competition,” The American Banker (July 7, 1993),
p. 17.

REFERENCES

Frankel, Alan S. “Monopoly and Competition in the Supply of Money
and Payment Services.” Memorandum, November 1995.

Andrew, A. Piatt. “Substitutes for Cash in the Panic of 1907,” Quarterly
Journal of Economics (August 1908). Reprinted in O. M. Sprague,
History of Crises under the National Banking System, U.S. National
Monetary Commission, Senate Document No. 538, 61 Congress, 2nd
Session, Government Printing Office, 1910, pp. 434-59.

Frodin, Joanna H. “Fed Pricing and the Check Collection Business:
The Private Sector Response,” Business Review, Federal Reserve Bank
of Philadelphia (January/February 1984), pp. 13-22.

FEDERAL RESERVE BANK

22

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Spahr, Walter E. The Clearing and Collection of Checks, The Bankers
Publishing Co., 1926.

Gilbert, R. Alton. “Utilization of Federal Reserve Bank Services by
Member Banks: Implications for the Costs and Benefits of
Membership,” this Review (August 1977), pp. 2-15.

Sprague, O.M. History of Crises under the National Banking System,
U.S. National Monetary Commission, Senate Document No. 538, 61
Cong. 2 Sess, Government Printing Office, 1910.

Gilbert, Richard J. “On the Delegation of Pricing Authority in Shared
Automatic Teller Machine Networks,” in Electronic Services Networks:
A Business and Public Policy Challenge, Margaret E. Guerin-Calvert
and Steven S. Wildman, eds., Praeger Publishing, 1991,
pp. 115-44.

Summers, Bruce J. “Risk Management in National Payment Systems,”
in The New Financial Landscape: Forces Shaping the Revolution in
Banking, Risk Management and Capital Markets, OECD (1995), pp.
253-280.

Horii, Akinari, and Bruce J. Summers. “Large-Value Transfer Systems,” in
The Payment System: Design, Management, and Supervision, Bruce
J. Summers ed., International Monetary Fund, 1994, pp. 73-88.

Veale, John M., and Robert W. Price. “Payment System Float and Float
Management,” in The Payment System: Design, Management, and
Supervision, Bruce J. Summers, ed., International Monetary Fund,
1994, pp. 145-63.

Jessup, Paul F. The Theory and Practice of Nonpar Banking,
Northwestern University Press, 1967.

Viswanathan, P., and Cesar Mayo. “A Note on ‘The Fed’s RCPC
Performance,’” Journal of Bank Research (Spring 1975), pp. 70-71.

Juncker, George R., Bruce J. Summers and Florence M. Young. “A Primer
on the Settlement of Payments in the United States,” Federal Reserve
Bulletin (November 1991), pp. 847-58.

Watkins, Leonard L. Bankers’ Balances, A.W. Shaw Company, 1929.

Katz, Michael L., and Carl Shapiro. “Systems Competition and Network
Effects,” Journal of Economic Perspectives (Spring 1994), pp. 93-115.

White, Eugene N. The Regulation and Reform of the American Banking
System, 1900-1929. Princeton University Press, 1983.

Marjanovic, Steven. “The Fed Buys Banctec Check-Image Storage
System,” The American Banker (February 1, 1996), p. 18.

White, Hubert D., and David A. Torgerson. “A Comment on ‘The Fed’s
Performance: What Does it Imply for EFTS?’” Journal of Bank
Research (Autumn 1974), pp. 193-96.

______. “Arizona Clearing House Chief Eyes Expansion,” The
American Banker (August 18, 1995a), p. 16.
______. “Chemical Joins ACH That Is Taking On the Fed,” The
American Banker (January 25, 1995b), p. 14.
______. “N.Y. Clearing House Sets Deadline for Electronic
Presentment,” The American Banker (March 23, 1995c), p.1.
______. “PEOPLE IN THE NEWS: Clearing House Group Names
President,” The American Banker (November 15, 1994a), p. 18.
______. “Citi, 4 Other Join Clearing House Group,” The American
Banker (December 7, 1994b), p. 16.
Morris, Russell D. “The Fed’s Regional Check Processing Performance:
What Does it Imply for Electronic Funds Transfer?” Journal of Bank
Research (Summer 1974), pp. 86-91.
______. “The Fed’s RCPC Performance: A Reply,” Journal of Bank
Research (Winter 1975a), pp. 257-59.
______. “A Response,” Journal of Bank Research (Spring 1975b),
pp. 72-73.
Richards, Heidi Willmann. “Daylight Overdraft Fees and the Federal
Reserve’s Payment System Risk Policy,” Federal Reserve Bulletin
(December 1995), pp. 1065-1077.
Roberds, William. “Financial Crises and The Payments System: Lessons
from the National Banking Era,” Economic Review, Federal Reserve
Bank of Atlanta (September/October 1995), pp. 15-31.
Salop, Steven C. “Deregulating Self-Regulated Shared ATM Networks,”
Economics of Innovation and New Technology, (1990), pp.
85-96.

FEDERAL RESERVE BANK

23

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Appendix

PAYMENT SERVICES OF THE
FEDERAL RESERVE BANKS

The Reserve Banks accept for
collection checks (cash letters) drawn on
banks located within the same territory, as
well as checks drawn on banks located in
other Federal Reserve territories.16 The
high degree of cooperation among the
Reserve Banks in processing interterritory
checks, and especially in debiting and
crediting reserve accounts of banks located
throughout the nation, facilitates an
efficient nationwide check-clearing and
settlement system.
The Reserve Banks accept shipments
of checks in various degrees of sorting,
including unsorted, sorted by Federal
Reserve territory, or sorted by the banks on
which the checks are drawn. Since sorting
checks by territories or by the banks on
which they are drawn is costly, the banks
that send checks to the Reserve Banks
already sorted are charged less than the
banks that send checks unsorted and rely
on the Federal Reserve Banks to sort them.
Like correspondent banks, Reserve
Banks credit the accounts of the depositors
of checks according to published availability schedules. That is, the depositors
are able to count on receiving credit for
checks drawn on banks located in different
Federal Reserve territories according to a
published time schedule, regardless of the
Federal Reserve’s ability to present the
items within that schedule. Any float that
results from mismatching the time of crediting the accounts of the depositing banks
and the time of presentment to and debiting the accounts of paying banks is a cost
of doing business for the Federal Reserve.
Float is factored into the base costs recovered through explicit fees.17
In recent years, the Reserve Banks
have provided a variety of value-added
check-clearing services, particularly
electronic information services demanded
by check-clearing customers. Banks that
receive cash letters (bundles of checks
written by depositors) from the Reserve

Cash Services
The Reserve Banks have a government
monopoly on issuing currency, which they
process for banks. They also process coin
issued by the United States Mint. Reserve
Banks process cash deposits and shipments
without charge to banks, at least for a
certain basic level of service, since this
is considered a government function.
In accepting deposits of currency and
meeting orders for currency, the Reserve
Banks maintain the quality and the
integrity of the currency stock. They
employ sophisticated processing equipment that separates currency fit for circulation from unfit currency, which is
destroyed. Reserve Banks also identify
counterfeit notes, relying on the anti-counterfeiting features built into the design of
the official currency.

Check Clearing

16

Cash letters are bundles of
checks accompanied by registers that list the contents of the
bundles and the total value of
the items they contain. These
bundles are called “cash letters” because settlement for
the checks is in cash-equivalent
funds, subject to the rules governing the return of checks.

17

For a discussion of how float
arises, is controlled, and what it
costs, see Veale and Price
(1994).

Since their formation, the Reserve
Banks have provided a nationwide check
clearing service. Check processing now
takes place at 11 of the 12 Reserve Bank
head offices, 24 of the 25 branches, 11
regional check-processing centers
(RCPCs), and one additional facility. The
Reserve Banks cooperate in managing this
system in a highly-integrated manner and
share some facilities and mechanisms,
including a national transportation system
for transporting checks.
Each Federal Reserve check-processing
office serves an official territory, which is
designated by routing numbers encoded at
the bottom of checks. A territory may
include a small but highly concentrated
area, such as a city, or a larger area with
banks dispersed across a large region.
Routing numbers are the de facto national
standard for check-clearing territories
throughout the United States.

FEDERAL RESERVE BANK

24

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Banks can receive electronic transmissions
that show the account numbers and dollar
values of individual checks, information
that normally would be physically presented later in the day. Timely availability
of this information greatly aids banks in
offering cash-management services to their
large corporate customers.
In addition, the Reserve Banks are
beginning to offer electronic check-deposit
and presentment services, as well as truncation services, for banks that elect to
settle checks in this manner. Reserve
Banks recently have begun to offer imageprocessing services both to commercial
banks and to the U.S. government. For
example, images of government checks are
captured and stored in archives, facilitating
federal government investigations and
claims settlements that involve payment by
check. In addition, commercial image services are now being provided to banks,
especially services designed to expedite
check adjustments and returns.

a large organization like the Federal
Reserve System, with its established technical infrastructure and its extensive
access to capital, was in a position to
invest in a venture of this magnitude, for
which the return was still uncertain. Moreover, the volume of payments at ACH’s
inception was not sufficient to justify the
costs of the large initial investment for
most companies. An exception was the
New York Clearing House, which chose
to handle the processing for ACH transactions in the Second Federal Reserve
District. Over the years, the Arizona
Clearing House and Visa USA, Inc. have
also established successful ACH processing
operations. The Chicago Clearing House
also attempted to offer an ACH processing
service but discontinued its service after a
few years.

Safekeeping of Definitive Securities
and Non-cash Collection
The Reserve Banks continue to
provide definitive safekeeping and noncash collection services to depository
institutions, but on a significantly reduced
scale in comparison to earlier years.18
Definitive securities are paper instruments,
such as bonds issued by state and local
governments. Safekeeping for such securities includes accepting them under a trust
agreement, collecting interest coupons,
and redeeming matured securities. Today,
only three Federal Reserve offices provide
these services as priced services, although
they do so for depository institutions
located throughout the United States.
Each Federal Reserve Bank does safekeep securities it accepts as collateral for
discount window loans and/or in its role
as fiscal agent. The demand for such services eventually will disappear completely
as all securities are converted from paper
to book-entry form, with ownership
recorded and transferred electronically in
the records of depositories.

Automated Clearinghouse
The automated clearinghouse (ACH)
is an electronic alternative to check
processing. In fact, the ACH was originally
most attractive as a means of converting
payroll disbursement and other recurring
money transfers from check-based transactions to electronic transactions. The ACH
is both a credit and a debit payment mechanism; that is, customer banks can make
payments and withdrawals from accounts
within the system. Use of ACH debit
entries increases the efficiency and speed
of transactions such as insurance
premiums and mortgage payments, and
they facilitate the concentration of cash by
treasurers of businesses that maintain
demand deposit accounts at large numbers
of banks.
The Reserve Banks began providing
ACH services to the United States Treasury
and commercial banks in 1972. The
federal government was a pioneer in converting its own paper-based check
payments to ACH. At its inception, ACH
processing required a very significant
investment in computer technology. Only

Wire Transfer of Funds and Securities
Banks can transfer reserves among
themselves electronically through the
Fedwire funds transfer service. This is a

FEDERAL RESERVE BANK

25

OF

S T. L O U I S

18

Non-cash items are handled on
a collection basis, meaning that
principle and interest are credited to the accounts of banks
with securities in safekeeping
when collected, not on the
basis of a published availability
schedule.

S E P T E M B E R /O C T O B E R 1996

19

The Reserve Banks maintain
book-entry securities accounts
for banks just as they maintain
funds accounts in which banks
hold reserves.

real-time gross settlement service that provides final payment. Each transfer is
processed separately without a netting of
payment messages among banks (gross
settlement), and the transfer of funds is
final, which means that it cannot be
reversed. If a bank sends a payment
message over Fedwire and later fails while
its reserve account is overdrawn, the
Reserve Bank holding the overdrawn
reserve account cannot recover funds from
the receiver of the payment message. The
finality of Fedwire funds transfers make
them a unique type of payment, quite distinct from other payment services provided by the Reserve Banks. For instance,
the Reserve Banks do not guarantee that
credits to reserve accounts resulting from
check collection are “good funds.” If a
Reserve Bank cannot collect from a bank
on which the checks have been drawn, it
has the right to reverse the credits to the
reserve account of the depositing bank.
Fedwire funds transfer is a natural
monopoly, in that only the Federal Reserve
can provide final settlement of reserves
transfers.
Reserve Banks began providing the
Fedwire funds transfer service in 1918 via
telegraph. Today the Reserve Banks
operate a highly-sophisticated computer
network with more than 8,000 on-line
connections to the Fedwire funds transfer
system. The Fedwire securities transfer
service dates to 1967, when the Reserve
Banks agreed with the United States Treasury to begin converting U.S. Government
securities to book-entry form. The
computer system of the Reserve Banks
became the depository for ownership of
the government securities. The Fedwire
securities transfer service is also a realtime gross settlement service, providing
for the simultaneous delivery of securities
and payment in final funds on the books of
the Reserve Banks.19 This delivery-versuspayment feature of securities transfers over
Fedwire limits the risk to participants in
the market for government securities,
because a seller of securities can transfer
ownership to a counterparty in a transaction without concern about whether the

counterparty will pay for the securities.
Since the Reserve Banks assume any
risk from the transfer of reserves over Fedwire, the Fedwire funds transfer service is,
together with federal deposit insurance
and the discount window, part of the federal safety net for the banking system. It’s
important to note that, given the high
volume, value, and velocity of wire transfers, the Fedwire service is able to operate
efficiently as a real-time gross settlement
system only because the banks that use
Fedwire have access to significant amounts
of intraday credit from the Reserve Banks.
When the Reserve Banks agree to process
Fedwire funds transfers on behalf of financially troubled institutions, they essentially
guarantee payments by these institutions,
thereby providing confidence to counterparties receiving the payments and contributing to the stability of the payments
system. The Reserve Banks manage the
risk in providing this guarantee through a
combination of operational and financial
controls.
An important milestone in the Fedwire
funds and securities transfer services was
the introduction of explicit pricing of Federal Reserve intraday overdrafts on April
14, 1993, to provide banks an incentive to
limit their use of intraday credit (Richards,
1995; Summers, 1995).

Fiscal Agency
One of the roles of the Federal Reserve
Banks is to serve as a fiscal agent for the
U.S. government. The Reserve Banks provide services to the United States Treasury
and to a variety of other government agencies, as requested by the Treasury
Department. For example, they collect
checks, process ACH transactions, and
make wire transfers on behalf of the fiscal
principals. They provide a variety of cash
management services for government
agencies, including collection, cash
concentration, and letters of credit. In
addition, they service the public debt and,
through the Fedwire securities transfer
service, provide operational support for
the secondary market in U.S. government
and agency securities. Approximately 12

FEDERAL RESERVE BANK

26

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

percent of the annual operating expense of
the twelve Reserve Banks is attributable
directly to their role as fiscal agents (Federal Reserve Planning and Control System).

provisional net settlement service represent “good funds.” If a Reserve Bank is
not able to collect the net debits from
members of a group, it may reverse the
entries made to the reserve accounts for
the net settlement.
One reason depository institutions
have a strong interest in using the Federal
Reserve for interbank settlement is that
virtually all depository institutions in the
United States hold reserve accounts at
the Reserve Banks (Blommestein and Summers, 1994). They are uniquely positioned
to meet the needs of clearinghouses with a
diverse membership, since virtually all the
clearinghouse members would hold
accounts with the Reserve Banks. In addition, the Reserve Banks are able to offer
their natural monopoly advantage of providing final settlement in central bank
money, rather than in terms of the
liabilities of another private bank.

Net Settlement
Net settlement is a service provided by
the Federal Reserve Banks to a group of
banks that clear payments among
themselves, net their interbank positions,
and settle their net debit and credit
positions through entries to their reserve
accounts. These arrangements can be classified as “final” or “provisional.” Net
settlement entries classified as final are not
reversible by the Reserve Banks, whereas
provisional entries are reversible.
In December of 1994, the Board of
Governors of the Federal Reserve System
issued a policy statement governing largevalue arrangements, which establishes the
conditions that such arrangements must
meet to operate and to gain access to Federal Reserve net settlement services.20
Large-value settlement arrangements
include electronic funds transfers (for
example, the Clearing House Interbank
Payments System or CHIPS) and electronic
securities clearing and settlement (for
example, the Government Securities
Clearing Corporation or GSCC). To minimize the systemic risk associated with
large-value netting arrangements, the new
policy statement requires that these
arrangements be designed to achieve final
settlement. Members use Fedwire to settle
their net debit obligations resulting from
net settlement. Also, one private ACH service and one check-clearing arrangement
use Fedwire for net settlement.
The Reserve Banks also provide settlement services for small-value payments to
approximately 160 local and regional
clearinghouses throughout the country.
The small-dollar settlements are for a
variety of netting arrangements, predominantly check clearinghouses, but also for
credit card systems and ATM and POS networks. Settlements for these netting
arrangements are provisional. The Federal
Reserve does not guarantee that credits to
reserve accounts resulting from use of its

20

FEDERAL RESERVE BANK

27

OF

S T. L O U I S

Federal Reserve Press Release,
December 21, 1994, Docket
No. R-0842.

S E P T E M B E R /O C T O B E R 1996

William T. Gavin is vice president and research coordinator for the Federal Reserve Bank of St. Louis. Dan Steiner provided research assistance.

The shaded box on page 30—“FOMC Discussion of Inflation Targets”—summarizes
these two discussions.
Those who advocate inflation targeting
argue that the FOMC would better communicate its intentions and thereby enhance
the credibility of its policy. Nevertheless,
some members are opposed to inflation
targeting, stating that such targets might
hinder the pursuit of effective countercyclical policies. In addition, some economists have argued that targeting the
aggregate price level directly would cause
instability in the economy and, possibly,
in the price level. This view is based on
the notion that prices adjust sluggishly
to all shocks. It gives little credence to a
rational expectations view of the monetary
transmission mechanism. The rational
expectations revolution in macroeconomic
theory, as well as the practical experience
of the United States and other countries in
setting monetary targets and attempting to
control inflation, has led many economists
(including this one) to favor using the
price level, rather than the money supply,
as a guide for policy.

The FOMC
in 1995:
A Step Closer
to Inflation
Targeting?
William T. Gavin
In the course of the Committee’s discussion of its monetary growth ranges, members
commented on the failure of the monetary
aggregates to provide a reliable nominal
anchor for the conduct of monetary policy
in recent years. Moreover, the restoration of
historic relationships, or the emergence of
new but stable relationships, between money
growth and measures of progress toward
broad economic objectives could not be
predicted with any degree of confidence.
Some members expressed the view that in
these circumstances the Committee needed
to continue to look at potential alternative
approaches to guide the formulation of
policy and to communicate its intentions to
the public, especially with respect to the
Committee’s objective of promoting price
stability over time.
Minutes of the FOMC meeting,
July 5–6, 1995, pp. 19–20.

Rational Expectations
Sargent (1986) explains how the
rational expectations revolution in macroeconomics changed the way economists
think about the monetary transmission
mechanism. Economists believe price level
is determined by expected future policies
rather than by past monetary policies. The
intuition behind rational expectations is
simply that people will use information
about the policy process when they form
expectations of inflation. This is an important insight because it suggests that the
long and variable lag from money to prices
is partly caused by changes in expectations
about future inflation that occur because
the policymakers and circumstances that
determine monetary policy change.1
News about monetary policy actions
is transmitted through all markets almost

The FOMC has long had a price
stability goal, but it has not accepted
suggestions to choose an index and set a
timetable that would make this goal an
operational target. The FOMC took up
the issue of inflation targeting at its
first meeting in 1995, discussing the
advantages and disadvantages of setting
explicit inflation targets. The issue also
came up at the Committee’s September
1995 meeting when it discussed the
Economic Growth and Price Stability
Act of 1995, a bill proposed in the Senate.

FEDERAL RESERVE BANK

29

OF

S T. L O U I S

1

See Bryan and Gavin (1994)
for details of this argument.
Gavin and Kydland (1995) present evidence of instability in
U.S. data, which, they argue,
can be explained by changes in
the monetary policy process.

S E P T E M B E R /O C T O B E R 1996

FOMC DISCUSSION OF INFLATION TARGETS
Setting Specific Inflation Targets

“The Committee also considered the potential advantages and disadvantages of
setting specific targets for bringing inflation down and achieving price stability over
time. Such targets might provide an alternative or supplemental approach to the monetary growth ranges, which had been found to be unreliable guides for monetary policy
over the past several years. The members discussed a number of aspects of inflation
targeting. On the one hand, such targeting would help to anchor the conduct of monetary policy and progress in meeting these objectives could enhance the credibility of
the Federal Reserve and perhaps reduce the overall cost of attaining price stability. On
the other hand, close adherence to preset inflation targets could unduly constrain the
Federal Reserve in its efforts to counteract the effects of cyclical shortfalls in the performance of the economy.”
Minutes of the FOMC Meeting, January 31–February 1, 1995, p. 18.
The Goal of Price Stability

“At this meeting, the Committee discussed a bill, titled the “Economic Growth and
Price Stability Act of 1995,” that recently had been introduced in the United States
Senate. The bill would make price stability the primary long-run policy goal of the
Federal Reserve and require the Federal Reserve to establish a numerical definition of
price stability and to implement a policy that would effectively promote such stability
over time. It would repeal the Full Employment and Balanced Growth Act of 1978 (the
“Humphrey-Hawkins Act”) and certain related provisions in the Employment Act of
1946 and the Congressional Budget Act of 1974. The Federal Reserve had not yet been
asked its views of the bill, but testimony was likely at some point and a preliminary
discussion would help to identify important issues.
“The members had not had time to review the bill in detail or to consider fully all
its implications. Nonetheless, their initial reaction was favorable in regard to the overall thrust of the bill’s monetary policy provisions. These would make clear that price
stability was the primary long-run objective of monetary policy and would restructure
the monetary policy reporting requirements to permit the Congress to carry out its
oversight responsibilities more effectively. Many members felt that in the context of
seeking and maintaining price stability, monetary policy should have the flexibility to
react to short-run fluctuations in output and employment, and they believed the bill
would be improved if its intent in this regard were clarified. A few members expressed
strong reservations about the part of the bill that would delete the employment objectives set forth in the Employment Act of 1946.”
Minutes of the FOMC Meeting, September 26, 1995, pp. 17-18.

2

See Litterman and Weiss
(1985) for an econometric
analysis of this view of the
monetary transmission mechanism.

simultaneously because changes in
expected monetary policy affect pricing
behavior in all markets. In markets where
contracts prevail, or where markets do not
meet and clear frequently, such monetary
policy changes will affect the distribution
of wealth even though the price changes
may not be recorded in transactions until
much later. With forward-looking expectations, the important channels through
which monetary policy affects real out-

comes are changes in the expected inflation trend and deviations of the actual
inflation rate from the rate that was
expected.2
The rational expectations revolution
implies that policies should be transparent.
Everything else equal, the policy process
should be structured to help people form
accurate expectations. For example, a
credible multiyear price-level objective
would help concentrate expectations about

FEDERAL RESERVE BANK

30

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

inflation. This, in turn, would cause
people to make decisions in a way that
would make the multiyear targets easier to
achieve.

Figure 1

M1 Velocity
Quarterly Data, Seasonally Adjusted
8.5

Experience with Monetary Targets

3% Growth Trend

8.0

Our understanding of monetary
phenomena has also been advanced
through two decades of experience with
monetary targeting in the United States.
The Monetary Control Act of 1980 was
structured under the assumption that
M1—the sum of currency in circulation,
traveler’s checks, and checkable deposits—
would be the key monetary target. This
decision reflected the relatively constant
growth rate of M1 velocity, the ratio of
aggregate income to the money stock. As
Figure 1 shows, M1 velocity appeared to
experience a steady 3 percent growth
during the 1960s and 1970s. The average
inflation rate was approximately equal to
the average M1 growth rate because the
real economy grew at the same average
rate as M1 velocity. The simplicity of the
long-run relationship of M1 to inflation
implied that policymakers could center the
M1 growth range on the desired inflation
trend.
Unfortunately, the simple relationship
did not hold after 1980. The introduction
of interest-bearing checkable accounts and
the implementation of disinflation policy
caused a dramatic change in the behavior
of M1 velocity. Rather than continuing to
grow at 3 percent per year, M1 velocity has
actually fallen during the past 15 years.
From 1981 through 1995, M1’s average
annual growth was just under 7 percent.
Under the simple rule of thumb that
worked before 1980, the average consumer
price index (CPI) inflation rate would
have been expected to be about 7 percent,
rather than the 4 percent average that was
actually observed. By 1987 the FOMC
stopped targeting M1 and emphasized the
broader aggregates.
Hallman, Porter, and Small (1991)
suggested that a long-run relationship
existed between M2 and the aggregate
price level. This relationship was reflected
in a relatively flat trend in M2 velocity.

7.5

M1 Velocity

7.0
Velocity Ratio

6.5
6.0
5.5
5.0
4.5
4.0
3.5
1960

1965

1970

1975

1980
Year

1985

1990

1995

2000

Figure 2

M2 Velocity
Quarterly Data, Seasonally Adjusted
2.04
M2 Velocity
1.96

Velocity Ratio

1.88
1.80
1.70 Average

1.72
1.64
1.56
1960

1965

1970

1975

1980
Year

1985

1990

Figure 2 shows M2 velocity with a trend
line calculated as the average from 1959
to 1989. M2 velocity displayed considerable cyclical variability and interestrate sensitivity, but Carlson (1989) and
Moore, Porter, and Small (1990) showed
that these fluctuations appeared to be
transitory and reasonably well explained
by standard money demand theory. The
apparent lack of trend in M2 velocity suggests that targets for M2 could be useful
in communicating the long-run inflation
objective. Here, the simple long-run
relationship suggests that nominal gross

FEDERAL RESERVE BANK

31

OF

S T. L O U I S

1995

2000

S E P T E M B E R /O C T O B E R 1996

with the length of the period during which
inflation was forecast: Forecasters could
predict the next quarter better than they
could forecast the average of the next two
years. Although inflation was relatively
predictable in the short run, the trend in
the inflation rate appeared to be highly
variable. Since 1990, it appears that this
longer-term uncertainty may be substantially reduced. Figure 3 depicts the actual
change in the CPI and the mean forecast
of inflation from the Survey of Consumer
Attitudes conducted monthly by the
University of Michigan Survey Research
Center. The mean value from this survey
is aligned so that the forecast matches the
period of actual inflation. Both the actual
and the expected inflation rate display
high year-to-year variability in the 1976–
1987 time frame McNees studied. Figure 3
also shows declining variability with the
decline in the average inflation rate after
1982.
Table 1 includes detail about the level
and variability of inflation for various subperiods and alternative inflation measures.
The average inflation level so far in the
1990s is about the same as it was in the
1960s, well below levels recorded in the
1970s and somewhat below those of the
1980s. In general, inflation volatility rose
with the average level of inflation, so we
should not be surprised to see volatility
in the 1990s below that observed in the
1980s. Nevertheless, it is difficult to compare inflation variability of the recent
period with the 1980s because the United
States has not had a recession or a major
oil price shock since the 1990–91 recession. Even with an optimal policy
process, we would expect inflation to
become more variable in response to such
disturbances.
As we might expect, inflation rate
stabilization has been associated with a
reduction in the uncertainty about inflation. Table 2 reports statistics from two
surveys of inflation expectations. The first
is the Michigan survey, also shown in
Figure 3; the second is the semi-annual
Livingston survey of economists that is
maintained by the Research Department at

Figure 3

Michigan Mean Expectations and Actual
CPI Inflation (Year-over-Year CPI)
Percent
15
12
9
6
3

Expected
Actual

0

60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98
Year
Note: The expected inflation in May 1997 is the forecast from May 1996.

domestic product (GDP) growth would be
equal to M2 growth. To achieve price stability, all the Fed would need to do is set
the M2 growth target equal to the trend in
real GDP growth.
From 1991 through 1995, M2 grew
at an average annual rate of 2.1 percent,
a little more than the average real GDP
growth rate of 1.9 percent. Ex ante, such
an outcome should have been expected to
have resulted in a stable price level—
an average inflation rate very near zero.
However, the relatively constant long-term
trend in M2 velocity also disappeared as
the economy recovered from the 1990–
1991 recession. There was a substantial
increase in M2 velocity, a surprisingly
stable 3 percent inflation trend, and a continuing de-emphasis of M2 in discussions
about the short-run stance of monetary
policy. This experience with M1 and M2
gives us less reason to agree that “the long
way around seems the surer way to our
objective” (Friedman, 1968, p. 15).

Controlling Inflation
The desirability of targeting inflation
has been affected by the FOMC’s success
in getting control over inflation in the past
15 years. In a study of errors in private
forecasts from 1976 through 1987,
McNees (1988) reported that uncertainty
about inflation went up, not just with the
length of the forecast horizon, but also

FEDERAL RESERVE BANK

32

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Table 1

Inflation Statistics
Period

CPI*

CPI(LFE)

PPI†

PPI(LFE)

PGDP‡

1961-70

3.0
(2.7)

3.2
(2.9)

1.8
(3.9)

Not
Available

2.9
(1.8)

1971-80

8.2
(4.5)

7.4
(4.2)

9.0
(8.5)

9.2
(5.7)

7.3
(2.3)

1981-90

4.6
(3.3)

5.1
(2.7)

3.0
(5.7)

3.6
(3.1)

4.3
(1.8)

1991-96

2.9
(1.3)

3.3
(1.2)

1.4
(3.3)

1.9
(2.7)

2.6
(0.8)

Note: Each cell contains the average inflation for the period and the standard deviation of the inflation rate in parentheses. All figures
are seasonally adjusted and expressed as compound annual growth rates.
* The CPI, the Consumer Price Index for all urban consumers, and the CPI less food and energy (LFE) are monthly.
† The PPI, the Producer Price Index for finished goods, and the PPI less food and energy (LFE) are monthly.
‡ PGDP, the Gross Domestic Product chain-type price index (1992=100), is quarterly.

Table 2

Inflation Forecast Errors
(Forecast Minus Actual)
Michigan Survey

Livingston Survey

Period

MAE*

Bias†

MAE

Bias

1961-70
1971-80
1981-90
1991-96

0.78
1.95
1.27
1.38

0.30
21.05
0.70
1.32

1.08
2.63
1.35
0.61

21.07
22.33
0.92
0.49

Note:
* MAE = Mean Absolute Error
†
Bias = Mean Error

the Federal Reserve Bank of Philadelphia.
Except for the case of the Michigan survey
in the most recent period, the surveys
show a positive correlation between the
average level and predictability of inflation.
In this case, the larger error is associated
with lower inflation, but this is almost
entirely a result of bias in the forecast.
Figure 4 shows the error in the mean
forecast by respondents to the Michigan
survey. The mean forecast has consistently
been either too high or too low. When
inflation was accelerating in the 1970s,
survey respondents were slow to catch on

to the policy process and thus tended to
underestimate the inflation rate. In the
early 1980s, when inflation was averaging
around 4 percent, it appears that people in
the Michigan survey did not really believe
inflation would remain at that low level.
Inflation forecasts were accurate from
1987 through 1990, a period of accelerating inflation. The bias in the errors is
quite pronounced over short periods. This
bias exists even today, when inflation is at
its lowest level since the early 1960s. The
inflation rate has stabilized at around 3
percent since 1990, but it does not appear

FEDERAL RESERVE BANK

33

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

ance against what these members viewed
as the principal risk in current circumstances—that of rising inflation.” The
FOMC issued an asymmetric directive at
its March 28, 1995, meeting to “provide a
clear signal of the Committee’s intention to
resist higher inflation” (minutes of FOMC
meeting, p. 23).
The statement of an inflation objective
also suggests that policy would become less
restrictive if there were no signs of accelerating inflation. The FOMC lowered the fed
funds rate to 5.75 percent in July 1995,
explaining, “The members agreed that
under present economic conditions a slight
easing of the stance of policy would incur
little risk of stimulating increased inflation
and would be entirely consistent with their
commitment to continued progress toward
price stability over time” (minutes of July
5–6, 1995, FOMC meeting, p. 20).
In December 1995 the Committee
explained its vote to lower the fed funds
target to 5.5 percent: “In any case, the
recent slowing of the economic expansion,
combined with the wage and price restraint evident at current levels of resource
utilization and continuing business efforts
to expand capacity, suggested that there
was little risk of a pickup in inflation”
(minutes of December 19, 1995, FOMC
meeting, p. 14).
This description of the policy objective
resembles the “opportunistic” view of
policy outlined by Orphanides and Wilcox
(1996) and by Orphanides, Small, Wieland
et al. (1996). This view is that the FOMC
should stabilize the inflation rate at the
recent trend (currently, about 3 percent in
the CPI) until some outside event such as
a favorable supply shock or an unavoidable recession causes a reduction in the
inflation rate. Then the FOMC would take
such an opportunity to lock in that lower
rate as its new objective. One attraction of
this approach is that the FOMC would
never intentionally engineer disinflation
that might cause a recession.
This opportunistic approach to policy
arises naturally in a committee setting
where the decision-making process
requires a compromise between members
whose views lie on a continuum.

Figure 4

Forecast Error in the Michigan Survey
Deviation of Expected from Actual Inflation
Percent
7.5
5.0
2.5
0
22.5
25.0
27.5

60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98
Year

that people who have responded to this
survey have had much confidence that
the trend would continue. The question
is whether the Federal Reserve could
enhance its credibility and reduce these
expectational errors by adopting explicit
inflation targets.
An interesting statement of the Federal
Reserve’s current long-term inflation objective can be found in the 1996 Monetary
Policy Objectives:
Most on the Committee anticipate
consumer price inflation at or somewhat below 3 percent in 1996. ... the
Committee recognized that its expectations for inflation do not imply that
price stability has yet been reached.
Nonetheless, keeping inflation from
rising significantly during economic
expansions will permit a gradual ratcheting down of inflation over the course
of successive business cycles that will
eventually result in the achievement of
price stability (p. 5).
This statement suggests that the objective is to keep inflation from rising above
the recent 3 percent trend through the end
of the current expansion. On page 20 of
the minutes from its January 31 – February 1, 1995, meeting, the FOMC explained
its decision to raise the fed funds rate
target from 5.5 percent to 6.0 percent on
February 1, 1995: “In terms of balancing
the policy risks that were involved, a
prompt move would provide some insur-

FEDERAL RESERVE BANK

34

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Table 3

1995 and 1996 Projections
FOMC Central Tendency Projections

Projections
for
1995
Projections
for
1996

Blue Chip Consensus Forecasts

Variable

July 1994

Feb 1995

July 1995

Feb 1995

July 1995

Nominal GDP
Real GDP
CPI
Unemployment Rate

5 -5 /
2 / -2 /
2 / -3 /
6 -6/

5-6
2-3
3-3 /
5/

4/ -4/
1/ -2
3/ -3/
5/ -6/

5.5
2.5
3.4
5.5

4.8
2.0
3.4
5.7

5.4
2.2
3.7
5.7

5.6
2.4
3.6
5.8

1

1

3

2

4

3

1

1

2

4

2

4

1

1

1

1

1

2

3

2

4

3

4

2

8

4

3

1

8

8

4/ -5 /
2 / -2 /
2 /-3 /
5/ -6/

Nominal GDP
Real GDP
CPI
Unemployment Rate

3

1

7

3

At one end of the continuum are
members who want to focus policy actions
on the short-run real economy and who
will be willing to resist signs of rising inflation pressures because such signs are
associated with high levels of aggregate
demand. Furthermore, this group may be
persuaded to adopt preemptive policies to
prevent the acceleration of inflation because they believe inflation is so costly to
eradicate. On the other hand, they will
resist attempts to lower the inflation trend
intentionally if they believe that doing so
would require higher unemployment and
slower growth.
At the other end of the continuum are
members who want a deliberate policy
with multi-year targets to eliminate inflation. They argue that announcing targets
in advance and implementing the policy
gradually will reduce any costs associated
with disinflation.
When inflation appears to be on the
rise, there is no need for compromise:
Everyone votes for monetary restraint.
But in times of steady inflation, the key
to compromise is members in the middle
of the continuum. They want price
stability but will wait for more favorable
circumstances.

4

4

8

4

3

3

1

1

8

4

4

8

jections reported to Congress in February
and July, pursuant to the HumphreyHawkins Act. The central tendency of
these forecasts is shown in Table 3.3
Blue Chip forecasts show the corresponding forecasts of private-sector
economists.
Both output and employment grew
more than expected in the second half of
1994. By the beginning of 1995, forecasters had therefore nudged up inflation
forecasts and reduced unemployment forecasts for the year. In February, the central
tendency of the FOMC’s forecast for output was almost perfectly centered on the
Blue Chip Consensus forecasts. The Fed
policymakers were predicting 3.0 percent
to 3.5 percent inflation in the CPI, while
the Blue Chip Consensus fell in the upper
part of that range.
As 1995 unfolded, recurring reports
of weakness in industrial production, the
leading indicators, retail sales, and employment growth led both the FOMC and the
private forecasters to revise their projections downward for real growth. By July
1995, the FOMC and nonvoting Federal
Reserve Bank presidents were a little more
pessimistic about real growth than were
the Blue Chip forecasters. On the inflation
side, both the Fed and Blue Chip had
essentially the same forecasts in July as
they had reported early in the year. The
Fed reported a smaller range, centered on
the same 3.25 percent midpoint. The Blue

1995 Economic Outlook: Turmoil
and Tranquility
FOMC members’ and nonvoting presidents’ forecasts are summarized in pro-

FEDERAL RESERVE BANK

35

OF

S T. L O U I S

3

Pakko (1995) provides a
detailed description of earlier
forecasts and policy actions in
1993–1994.

S E P T E M B E R /O C T O B E R 1996

Figure 5

Blue Chip Forecasts for 1995
(Fourth Quarter/Fourth Quarter)
Nominal GDP

Real GDP

Percent
6.0

Percent
3.0
2.8

5.5

2.6
2.4

5.0

2.2
2.0

4.5

1.8
4.0

J

F

M

A

M

J
J
Month

A

S

O

N

D

1.6

J

F

M

CPI

Unemployment Rate*

Percent
4.0

Percent
6.0

A

M

J
J
Month

A

S

O

N

D

A

M

J
J
Month

A

S

O

N

D

5.8

3.5

5.6
3.0
5.4
2.5
2.0

5.2

J

F

M

A

M

J
J
Month
* Forecast for the average unemployment rate in 1995: 4

A

S

O

N

D

5.0

J

Chip reported the same 3.4 percent consensus figure as in February.
The monthly evolution of the Blue
Chip forecasts for growth in nominal
GDP, real GDP, and the CPI are shown in
Figure 5. Also included are the monthly
updates to the forecast for the fourth
quarter average unemployment rate. Two
important trends appeared in the evolving
outlook for 1995—the general decline
in inflation expectations that occurred
during the second half of the year and
the U-shaped pattern in the outlook for
growth in the real economy. By midyear
many economists were asserting that a
recession had begun. Yet the second half
of 1995 turned out to be surprisingly
robust, with output bouncing back so

F

M

strongly that the year actually finished
above the initial forecasts (when measured by the 1987 fixed-weight output
measure that was the basis for many
beginning-of-the-year projections).
The monetary aggregates are shown in
Figure 6. The introduction of sweep accounts caused M1 to decline throughout
most of 1995. New software made it easy
for depository institutions to sweep funds
automatically out of checkable accounts
into savings accounts to avoid the reserve
requirement on checkable deposits. Funds
would be moved back into the checkable accounts when needed to make
payments. Since these funds were swept
into accounts included in M2, this broad
aggregate was not directly affected by

FEDERAL RESERVE BANK

36

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Figure 6

Monetary Aggregates with Targets
(Monthly Data, Seasonally Adjusted)
M1

M2

Billions of Dollars
1160

Billions of Dollars
3720
3680

5%

1150
3640
1140

3600
3560

1130
1%

3520
1120

3480
1994

1994

1995

M3

Domestic Nonfinancial Debt

Billions of Dollars
4600

Billions of Dollars
14100

1995

7%

13900
4500

13700

6%

13500
3%

13300

4400
2%

13100

4300

12900
12700

4200

12500
1994

1994

1995

1995

Note: The M3 target was set at 0 percent to 4 percent and was changed to 2 percent to 6 percent after the July 5-6, 1995, FOMC meeting.
Table 4

Monetary Policy Objectives
Percentage Growth Rates
Meeting

Target Period

M2

M3

Debt

January 31-February 1, 1995

1994:Q4-1995:Q4

1-5

0-4

3-7

July 5-6, 1995

1994:Q4-1995:Q4

1-5

2-6

3-7

1995:Q4-1996:Q4

1-5

2-6

3-7

sweep activity. The targets for the broad
monetary aggregates are shown in Table 4,
as well as in Figure 6. M2 grew in the
upper part of its 1 percent to 5 percent
growth range. From early in the year, M3
grew above its initial 0 percent to 4 percent

target range. In July, the FOMC raised this
target range to 2 percent to 6 percent. The
monitoring range for nonfinancial debt
was set at 3 percent to 7 percent for 1995.
This measure of debt grew in the upper
part of the range throughout the year.

FEDERAL RESERVE BANK

37

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

exception of this temporary increase in
July, 1995 was a year in which the level of
rates fell and the term structure flattened.
The bottom panel of Figure 7 shows
the target level for the fed funds rate, again
accompanied by the actual fed funds rate
and the discount rate. The variability of
the actual fed funds rate demonstrates that
the Fed does not control the fed funds rate
directly but, rather, supplies reserves in a
way that keeps the average trading range
very close to the target rate. The upward
move to 6 percent in February, followed
five months later by a decline to 5.75 percent, is shown in the figure. On December
19, 1995, the rate was lowered again to 5.5
percent. The discount rate had been raised
to 5.25 percent in February and held at
that level through the end of the year.

Figure 7

Selected Interest Rates
(Daily Data, Annual Percentage Rates)
Percent
8.5
8.0
30 Year T-Bill
7.5
7.0
10 Year T-Bill
6.5
6.0
1 Year T-Bill
5.5
5.0
4.5
4.0
Fed Funds Target
3.5
3.0
2.5
J F M A M J J A S O N D J F M A M J J A S O N D J
1994
1995
Month
Percent
8.5
8.0
7.5
7.0
6.5
Fed Fund's Rate
6.0
5.5
5.0
Discount Rate
4.5
4.0
3.5
3.0
2.5
J F M A M J J A S O N D J F M A M J J A S O N D J
1994
1995
Month

FOMC DISCUSSIONS AND
DECISIONS
Several common themes emerged in
discussions on the outlook for the economy at all the FOMC meetings:
•The deceleration in the growth of
nominal hourly compensation. This
slowing in nominal labor costs spanned
most major occupations and industries
and was believed to reflect a fundamental
lowering of inflation expectations.
•The view that inventories were above
desired levels in many sectors and that, on
average, 1995 would be a year in which
attempts to reduce inventory accumulation
would lead to cuts in production and
slower real GDP growth.
•The notion that business fixed investment would stop growing so rapidly and
slow down to a more sustainable rate.
•The view that, by the end of 1994,
the economy had fully recovered from the
last recession and that pent-up demand for
housing, autos, and other durable goods
was exhausted.
The Committee’s decisions are summarized in Table 5. The directive to the
Manager of the Open Market Desk
contains language indicating the FOMC’s
intentions about a possible policy change
within the intermeeting period, depending
on developments in that period. Table 5

The top panel of Figure 7 depicts the
federal funds rate target and the evolution
of three interest rates: yields on 1-year,
10-year, and 30-year, constant-maturity
Treasury securities. Aggressive 1994 policy
actions and a 50 basis-point increase in the
fed funds target at the first FOMC meeting
of 1995 quelled expectations of rising
inflation and led to a year-long bond rally
as surveys and market measures of
inflation expectations inched downward
throughout the year. An interruption in
the trend occured in July 1995, after the
FOMC lowered the fed funds rate on the
day before the Bureau of Labor Statistics
announced a large increase in jobs for
June and a major upward revision of
employment statistics for May. Subsequent
economic reports showed that inflation
pressures were easing even as the economy
appeared to be strengthening. With the

FEDERAL RESERVE BANK

38

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Table 5

FOMC Directives and Measures of Monetary Policy Stance
1995
Meetings

Directive
for
Pressure

Intermeeting Stance
toward
Lesser
Greater
Restraint
Restraint

Jan. 31-Feb. 1

Increase
somewhat

Would

Mar. 28

Maintain (+)

May 23

Result from Change in Reserve Pressure
Date of
Change

Fed Funds
Target*

Discount
Rate

Would

Feb. 1

6.00

5.25

Might

Would

N/A†

6.00

5.25

Maintain

Would

Would

N/A

6.00

5.25

Jul. 5-6

Decrease
slightly (2)

Would

Might

Jul. 6

5.75

5.25

Aug. 22

Maintain

Would

Would

N/A

5.75

5.25

Sep. 26

Maintain

Would

Would

N/A

5.75

5.25

Nov. 15

Maintain

Would

Would

N/A

5.75

5.25

Dec. 19

Decrease
slightly

Would

Would

Dec. 19

5.50

5.25

* Federal funds rate expected to be consistent with desired reserve restraint. The rate expected to be consistent with policy before the
Jan. 31-Feb. 1 meeting was 5.5 percent. The discount rate was raised from 4.75 percent to 5.25 percent on Feb. 1. An asymmetric
policy setting is indicated by a (+) or (2) in the second column.
† N/A indicates that there was no change in the discount rate or the fed funds target.

prisingly robust labor market report on
July 7 and subsequent positive reports on
economic activity. The switch in beliefs
about real growth caused bond prices to
decrease temporarily. Throughout the
summer and fall, news continued to show
a robust economy and declining inflation.
The bond market turned around again
with the news of lower inflation and the
rally dating from the final months of 1994
continued through 1995. Falling long-term
interest rates and inflation expectations,
combined with some year-end slowing in
the economy, led the FOMC to lower the
fed funds target to 5.50 percent at the
December 19th meeting. I present details
about the decisions and discussions in the
following accounts of each meeting.

shows that, at the first meeting of 1995,
the Committee used the same word,
“would,” indicating that either greater or
lesser restraint would be acceptable within
the intermeeting period. An example of an
asymmetric directive was the one written
after the March 28, 1995, meeting, in
which the Committee used the weaker
word “might” to indicate that an easier
policy was less likely to be acceptable than
a tighter policy.
The FOMC began the year concerned
about the inflationary pressures implied by
the relatively high demand observed in
1994, and those pressures began to show
through in reports of consumer and
producer prices. Incoming data showed
that the economy was weaker than had
been expected in early 1995. In May the
FOMC adopted a symmetric directive,
switching from a position of prospective
tightening to a more neutral posture. Negative economic reports continued, leading
the FOMC to reduce the target for the fed
funds rate to 5.75 percent on July 6.
Indicators of a slowing real expansion appeared to turn around with a sur-

January 31–February 1 Meeting

The year began with the federal funds
rate trading at 5.5 percent. The Open
Market Desk operated under an asymmetric directive that called for a bias
toward a firming of reserve conditions
during the intermeeting period. The most
important economic news that had been

FEDERAL RESERVE BANK

39

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

released since the December 20, 1994,
FOMC meeting was a report showing a
surge in the fourth quarter GDP to an
annual rate of 4.7 percent. Although rapid
accumulation of inventories had been a big
part of the growth, analysts were divided
about whether this accumulation had
resulted from falling demand or from anticipation of future sales. News reviewed at
the meeting suggested that retail sales were
mixed but that consumer confidence was
high, and strength in the housing sector
was expected to support moderate growth
in 1995, even in the face of a probable
inventory correction. Reports about the
economy in the closing months of 1994
indicated that both industrial production
and payroll employment continued to grow
above their longer-term trends.
Financial markets were embroiled in
concerns about debt quality in the wake of
the Orange County, California, default and
the crisis in Mexico. Neither the FOMC
nor the markets were giving much weight
to monetary aggregates’ sluggish 1994
growth. Although M3 had picked up with
the rising demand for bank loans at the
end of 1994, M2 had grown along the
bottom of the 1 percent to 5 percent target
range. Total non-financial debt was
increasing slightly below the trend growth
in nominal spending. Although rising
money market interest rates in 1994 had
brought a halt to growth in M1, the Committee saw little indication that credit
conditions were tight.
Household and economic forecaster
surveys all predicted an acceleration of
inflation in 1995. To reduce the probability
and extent of such an acceleration, the
FOMC unanimously voted to increase
reserve market pressures so that the fed
funds rate would trade around 6 percent.
In conjunction with this decision, the Board
of Governors voted to raise the discount
rate from 4.75 percent to 5.25 percent. The
action was explained in a press release:
Despite tentative signs of some
moderation in growth, economic activity has continued to advance at a
substantial pace, while resource utilization has risen further. In these circum-

stances, the Federal Reserve views
these actions as necessary to keep
inflation contained, and thereby foster
sustainable economic growth. (Feb. 1,
1995.)
This 50–basis point increase in the fed
funds rate was followed by lower interest
rates in the bond market during the intermeeting period, as shown in Figure 7.

Crisis in Mexico
At the Committee’s first meeting of
1995, members also voted to expand the
swap arrangement with the Bank of Mexico:
On December 30, 1994, the Committee approved a temporary increase
from $3 billion to $4.5 billion in the
system’s reciprocal currency (swap)
agreement with the Bank of Mexico
and it also approved the activation of
that agreement. The Committee approved a further temporary increase of
$1.5 billion and activation of that
amount at this meeting, thereby raising
the swap arrangement with the Bank of
Mexico to a level of $6 billion, consisting of the regular $3 billion line and a
special $3 billion line. (Minutes of
FOMC meeting, p. 24.)
In addition, the Committee voted to
facilitate U.S. participation in the Multilateral Program to Restore Financial Stability
in Mexico, with an increase in the agreement to “warehouse” foreign currencies
for the U.S. Treasury:
The Committee also approved at
this meeting an increase from $5 billion
to $20 billion in the amount of eligible
foreign currencies that the System is prepared to ‘warehouse’ for the Treasury and
the Exchange Stabilization Fund (ESF)”
(Minutes of FOMC meeting, p. 25).
The majority of members voted to
approve these measures because they
“were persuaded that the nature and
severity of Mexico’s financial problems
could not be contained without making
available substantial financial assistance to
the Government of Mexico” (minutes of
FOMC meeting, p. 26). Two members
dissented:

FEDERAL RESERVE BANK

40

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

expected the FOMC to make further progress toward price stability. Few indications of monetary restraint surfaced as
bank credit continued to advance and the
broad monetary aggregates appeared to be
growing within the prescribed ranges.
In sum, there was a consensus among
Committee members that the economy
was slowing from the torrid pace of 1994
but was already operating at a high level,
and the extent of the slowdown was in
question. The Committee voted unanimously to maintain the degree of the
reserve restraint (fed funds trading at
about 6 percent), but concern about the
risk of accelerating inflation and a desire to
move gradually toward price stability led
them to adopt an asymmetric directive
biased in the direction of more restraint.

Messrs. Lindsey and Melzer dissented with respect to increases in both
the swap line and the warehousing arrangement with the Exchange Stabilization Fund. They did not believe that
the Committee had been provided sufficient information to assess whether
developments in Mexico threatened
U.S. financial stability, a possible justification for increased central bank
lending on a short-term basis. Furthermore, they considered it inappropriate
for the Federal Reserve to participate,
directly or indirectly, in intermediate to
long-term financing to facilitate debt
restructuring. They were concerned
that such participation in a fiscal policy
matter might compromise, or appear to
compromise, the independence of the
monetary policy process. Mr. Lindsey
added that the latter risks were significantly enhanced given the absence of
Congressional authorization or more
general public support for these measures. (Minutes of FOMC meeting,
pp. 26–27.)

May 23 Meeting

By the time of the May meeting, the
Committee was picking up signs of weakness in the year’s first half, particularly in
the demand for autos and housing. This
perception was reinforced by reports that
indicated slower job growth, a jump in the
unemployment rate to 5.8 percent in April,
a slowdown in the manufacturing sector, a
decline in retail sales, and a weakness in
export demand associated with problems
in Mexico. These indicators of weakness
were offset by expectations that the economy, especially the interest-sensitive sectors, would benefit from the ongoing rally
in bond markets and the surge in stock
prices. The rising stock market was reducing the cost of capital to businesses and
enhancing the wealth of many households.
Declining long-term interest rates were
accompanied by declines in survey measures of inflation expectations despite the
above-average growth in the CPI and the
producer price index (PPI) in the first four
months of the year. Monetary aggregates
grew little, and the market seemed to
believe the increase in inflation would be
transitory—the Blue Chip consensus remained at 3.4 percent.
Overall, the perceptions of weakness
in demand led the Committee to revise its
view that the next move in interest rates

March 28 Meeting

Given the rapid economic growth in
1994, policymakers anticipated a relative
slowdown in 1995. Early reports on economic activity were mixed. Industrial production continued to grow, and new jobs were
created at a rate that exceeded the underlying trend in labor force growth. The
expected inventory correction was not yet
evident; available information showed that
inventories had continued to accumulate in
January. On a weaker note, reports of spending in the first two months of 1995 showed
a decline in auto sales and a slowdown in
the housing market. Lower demand for
autos led to cutbacks in production schedules and this, combined with a bleak
houing market, led Blue Chip economists to
nudge GDP forecasts down slightly in 1995.
Expectations for a slight pickup in
inflation were reinforced as reports for the
first two months showed both consumer
and producer prices rising at rates above
their 1994 averages. Although private forecasters were lowering predictions of inflation slightly, there was no sign that anyone

FEDERAL RESERVE BANK

41

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

quarters were to continue, a 1 to 5 percent range for growth of M2 likely
would prove consistent with the Committee’s ultimate objectives of sustained economic expansion and reasonable price stability (Minutes of FOMC
meeting, p. 16).
Not all the members agreed that the
1 percent to 5 percent growth range was
appropriate for M2:
Mr. Blinder and Ms. Yellen dissented on a technical judgment, not a
policy difference. They noted that if
growth in the demand for M2 were
close to historic norms in 1995 or
1996, as indeed it had been for some
time, then the Committee members’
projections for nominal GDP would
likely imply M2 growth near the top
of, or even above, the current range.
While the relationship between the
growth of M2 and that of nominal GDP
remained subject to a great deal of
uncertainty, they were persuaded that
the range—in fact, the midpoint of the
range—should normally be consistent
with members’ forecasts of nominal
GDP growth. (Minutes of FOMC
meeting, p. 18.)
Strong growth in loans stimulated
bank demand for M3 liabilities and caused
M3 to grow above the upper 4 percent
limit. Projections of continued growth in
M3 led the FOMC to raise the target for
this broad aggregate by 2 percentage
points. The Committee voted unanimously
to raise the M3 target to 2 percent to 6
percent and to reestablish the monitoring
range for domestic nonfinancial credit at 3
percent to 7 percent. The Committee also
voted unanimously to adopt tentative targets for 1996 that were the same as the
targets agreed to for 1995.
In spite of reports of higher inflation
early in the year, the Committee left its
inflation outlook centered on 3.25 percent.
Blue Chip forecasters ignored these reports
as well, upholding their January inflation
outlook of 3.4 percent. The May CPI
report showed some slowing from the
early data. The flat PPI for finished goods
in May suggested some reduction of infla-

would be upward. It unanimously set aside
the asymmetric directive of the previous
meeting and voted in favor of a symmetric
policy that would maintain the existing
degree of pressure on reserve markets.
July 5–6 Meeting

Committee members’ differing perceptions had a strong influence at this midyear policy meeting. Some members
believed the economy would be substantially weaker than expected. On June 2,
the Bureau of Labor Statistics reported a
101,000 decline in the number of nonfarm
payroll jobs in May. This was after a slight
decline in April. The household surveys,
although thought to be less reliable, showed
even greater declines in April and May.
This unexpected weakness in the labor
market led Blue Chip forecasters to revise
their estimate of GDP growth for 1995
substantially downward. The consensus
forecast fell from 2.4 percent in the May
release to 2.2 percent in the June 10th
report. In July, Blue Chip forecasters again
lowered the 1995 outlook for GDP, this
time from 2.2 percent to 2.0 percent.
Although these figures were not released
until after the July FOMC meeting, most
of the individual forecasts would have
been made with the same information
available to the FOMC members who had
reduced their forecasts from February. The
central tendency in the outlook for real
GDP growth in 1995 fell from a range of
2 percent to 3 percent in February to 1.5
percent to 2.0 percent in early July. Pessimism brought on by the dramatic decline
in jobs was reinforced by the string of negative reports on retail sales, industrial
production, and the leading indicators.
M2 rebounded strongly after April and
approached the upper 5 percent limit of
the target range by midyear. In choosing to
keep the 1 percent to 5 percent range for
M2, a majority of the members reasoned
that they did not have enough confidence
in the relationship between measures of
economic performance and M2 to justify
making a change in the target:
Moreover, if the more normal behavior of velocity over the past several

FEDERAL RESERVE BANK

42

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

erate expansion. Revised data showed that
retail sales had not been as weak as reported earlier. The housing market picked
up considerably in response to falling
long-term interest rates. Business investment, led by spending on computing
equipment and construction, continued to
post solid gains. Although the state of the
economy remained uncertain, the news
was more balanced than it had been earlier
in the year. The good news about the
economy caused a temporary backup in
interest rates as market participants realized that aggregate demand was stronger
than previously thought.
CPI inflation, after rising at a 3.25 percent rate through the first five months of
the year, rose at more subdued rates in
June and July—the two months averaged
1.2 percent at an annual rate. Increases in
producer prices at the earlier stages of
production appeared to be decelerating,
suggesting some further reduction of
inflation later in the year. An ongoing
slowdown in the growth of benefit costs
supported continued moderate growth in
labor costs. By early August, Blue Chip
economists were beginning to mark down
their forecasts for CPI inflation in 1995.
The wedge between growth in the
narrow and broad monetary aggregates
increased. The narrow aggregates, from
currency to M1, were dramatically below
expected trends. The slowdown in currency was attributed to an unexplained
decline in the net foreign demand for U.S.
currency. The slowdown in the deposit
component of M1 was attributed to the
introduction of sweep accounts that
substantially reduced banks’ required
reserves.
The broad aggregates grew well above
the trend that had been established in
recent years. Some observers noted that,
following a period of rising velocity, M2
velocity appeared to be stabilizing around
a new higher level. M3 growth was also
rapid, reflecting the continued growth in
managed liabilities needed to fund strong
growth in bank loans.
This good news about the economy
led the Committee to vote unanimously to

tion pressures in the second half of the
year. At this meeting, members expressed
some differences of opinion about the definition of price stability. Some members
believed the weakening economy would
prevent an acceleration of inflation—and
therefore policy should be eased substantially. Other members stated that there was
no indication of progress toward price stability—and therefore policy should be held
at a 6 percent fed funds rate. The Committee compromised by slightly easing
reserve pressures. In the press release
issued at the end of the meeting, on July
6th, the Fed explained:
As a result of the monetary tightening initiated in early 1994, inflationary pressures have receded enough to
accommodate a modest adjustment in
monetary conditions.
Today’s action will be reflected in a
25 basis point decline in the federal
funds rate from about 6 percent to
about 5.75 percent.
There was one dissent on the shortrun policy decision:
Mr. Hoenig dissented because he
believed the stance of monetary policy
should remain unchanged at this time.
With the pace of economic activity
likely to return to trend growth later
this year and inflation expected to be
higher this year and next than in 1994,
he felt an unchanged policy in the near
term would enhance the prospects of
achieving the Committee’s long-run
objectives of sustainable economic
growth and price stability. (Minutes of
FOMC meeting, p. 24.)
August 22 Meeting

The news leading up to the July
meeting had been almost entirely negative.
But one day after the July meeting ended,
the monthly labor report showed an unexpected 215,000 increase in nonfarm
payroll jobs during the month of June; it
also revised the earlier report of a 101,000
jobs decline for May to show a decline of
only 46,000 jobs. Incoming information
suggested that the economy, after a weak
first half, was on track for continued mod-

FEDERAL RESERVE BANK

43

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

maintain the existing degree of pressure
in the reserve market—a fed funds rate
trading around 5.75 percent—and to abandon the bias in favor of an easier policy
that had been adopted at the previous
meeting.

degree of pressure in reserve markets,
with no bias for a change before the next
meeting. Some members were concerned
about downside risks to the economy and
believed the 5.75 percent fed funds target
was slightly restrictive.
However, the current performance
of the economy suggested that the
timing of an easing action was not an
immediate concern. Other members
who preferred an unchanged policy
placed more emphasis on current forecasts of little or no progress in reducing inflation from recent levels.
They thought it would be premature to
ease policy without greater assurance
that inflation had been contained in
the current cyclical expansion and that
prospects for significant further progress toward the long-run objective of
price level stability had improved.
(Minutes of FOMC meeting, p. 14.)

September 26 Meeting

Good news of lower inflation and the
real economy’s continued expansion filled
most reports during the intermeeting period. In September, the Blue Chip consensus forecast slightly higher spending in
1995, with a bit more real output growth
and a bit less inflation. Indicators that had
caused concern about economic weakness
early in the year turned around during the
summer:
•Consumer spending continued at a
relatively high level.
•Housing markets strengthened in the
presence of lower mortgage rates.
•The manufacturing sector surged in
August, led by a sharp increase in
auto production.
•Growth in employment rebounded
strongly from slower growth in July.
•Business spending on equipment
and structures remained surprisingly
robust.
News about consumer and producer
price inflation suggested that acceleration
in the first six months was likely to be
reversed in the second six months. On a
year-over-year basis, inflation was moderate in 1995, approximately equal to or
below that of the previous year and the
previous five years. The broad monetary
aggregates continued to grow in the top
part of the range (M2) or well above the
target range (M3). The term structure of
interest rates, while still above the low
level reached at midyear, had declined
slightly across the maturity spectrum from
the levels at the time of the August meeting.
Overall, information available to the
Committee at this meeting indicated continued strength in aggregate demand and
a moderating inflation trend, suggesting
little reason to adjust the degree of pressure in reserve markets. Members voted
unanimously to maintain the existing

November 15 Meeting

Information released before this
meeting revealed a surge of economic
growth in the third quarter that caused
forecasters to revise the GDP outlook for
1995 substantially upward. The Blue Chip
consensus for real GDP growth in 1995
was 2.7 percent in the November report,
up from 2.2 percent a month earlier. As
often occurs in the face of a surprisingly
good report for the previous period, the
GDP outlook for the current quarter was
revised downward. Falling retail sales in
October, slower employment growth in
September and October, falling industrial
output with a labor strike in the aircraft
industry, and uncertainty about the duration of the government shutdown led some
members to question whether aggregate
demand would be sustained, given the
current policy setting. Other members
believed there was sufficient evidence of
continued expansion to delay any policy
easing. Anecdotal evidence suggested that
retail sales were beginning to rebound in
early November; a strong housing market
earlier in the year was expected to fuel the
demand for household appliances and
other durable goods; and growth in

FEDERAL RESERVE BANK

44

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

business investment, although slowing
somewhat, remained on an upward
trend.
On the price side, reports of lower
inflation caused analysts to revise expectations downward. Long-term bond rates fell
with inflation expectations, and the Blue
Chip forecasters continued the march
downward in their outlook for CPI inflation for both 1995 and 1996. On the short
end of the market, the yield curve became
U-shaped as the one-year rate fell below
both the three-month and the three-year
rates. Growth in the broad monetary
aggregates, at or above the upper limits of
their respective target ranges, slowed in
October as the demand for bank loans
slowed. All these factors pointed toward
lower inflation pressures. As reported in
the minutes of the meeting, “With regard
to inflation, members noted that despite
generally high levels of resource use,
including tight labor markets in many
parts of the country, inflation had been
more subdued than many had expected
over the past several months.” Although
inflation expectations were lowered, they
still remained at or above the actual inflation trend in the economy.
Deliberations at the November meeting reflected both differences in views
about the outlook and differences about
the strategy that might be taken to achieve
price stability. The majority of the Committee agreed that aggregate demand was
sufficiently high to justify leaving the
trading range for the fed funds rate at
5.75 percent despite considerable sympathy with the view reflected in the term
structure—that is, during the next year,
short-term interest rates would move
lower. Most members who would have
preferred to cut rates at this meeting were
willing to wait for further information
about the state of the economy and the
associated demands for credit. The
meeting’s minutes explicitly state that
monetary policy should not be conditioned directly on the budget negotiations,
but rather, monetary policy should deal
directly with any change in the net demand
for credit that resulted from a budget deal.

The Committee voted to maintain the
existing degree of pressure on reserve markets. Even those who might have preferred
a bias toward a lower interest rate agreed
that there was not likely to be enough new
information available to justify an intermeeting policy adjustment. Only Governor
Lawrence B. Lindsey dissented from the
majority opinion:
Mr. Lindsey dissented because he
believed that monetary policy should
be eased. The evidence suggested to
him that in the absence of an easing
move the underlying rate of nominal
GDP growth was likely to be lower
than needed to maintain real GDP at
or near its potential. The intermediate
forecast was subject to a number of
significant risks: household balance
sheets seemed unlikely to sustain the
current rate of durables expenditure
for any extended period; government
expenditures were certain to be cut
substantially; and with fiscal
contractions underway in Europe and
Canada and severe financial stresses
present in Japan and Mexico, he
did not see much likelihood of a
substantial expansion of exports. In
keeping with his views, the financial
markets were signaling the likelihood
that a weaker pace of nominal GDP
growth would materialize. The
yield curve was virtually flat, with
government securities up through relatively long maturities trading at yields
below the current average federal funds
rate. Thus, markets would be unlikely
to find some easing inappropriate and
over the intermediate horizon would
view the current level of short-term
rates as unsustainable. (Minutes of
FOMC meeting, p. 18.)
December 19 Meeting

As FOMC members prepared for their
final meeting of the year, financial markets
were reflecting a continuing decline in
inflation expectations, and Committee
members were somewhat concerned that
fourth-quarter aggregate demand was
slightly weaker than had been anticipated

FEDERAL RESERVE BANK

45

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

a press release dated December 19, 1995,
Fed Chairman Alan Greenspan announced
that the easing was made possible because
“inflation has been somewhat more favorable than anticipated, and this result along
with an associated moderation in inflation
expectations warrants a modest easing in
monetary conditions.”

at the November meeting. The Blue Chip
consensus released in December forecast a
slight modification in the 1995 real GDP
growth rate to 2.6 percent, down 0.1 from
November’s report. This lower forecast for
the year was based on a view that the
fourth quarter would see little increase in
manufacturing, the tangled budget process
would result in a temporary reduction in
federal outlays, and both consumer and
business demand for current output would
moderate from earlier expectations. The
economy was expected to grow closer to
its perceived trend rate—around 2.5
percent in the unchained measure, down
from the rapid growth in the third quarter.
Consumer demand was thought to be
restrained by job insecurity associated
with widespread reports of business
restructuring; higher debt service burdens;
and the satisfaction of pent-up demands
for housing, autos, and other durable goods.
Growth in business fixed investment was
expected to slow just because few economists believed it could continue growing
as rapidly as it had done since 1992.
The reports on inflation reinforced a
widespread impression that 1995 would be
another year in which the CPI would
increase less than 3 percent. These lower
inflation expectations were evident in surveys, as well as in financial markets, where
interest rates continued to fall across the
entire term structure. The growth in the
broad aggregates remained subdued in
November after experiencing sluggish
growth in October. Nevertheless, robust
growth earlier in the year meant that M2
finished the year in the upper portion of its
targeted 1 percent to 5 percent range, and
M3 finished the year above the upper end
of its targeted 2 percent to 6 percent range.
Falling inflation and, more important,
expectations for lower inflation in the
future, as well as some indications of slowing in the expansion, led the Committee to
vote unanimously for a slight decrease in
reserve pressure, lowering the target for
the fed funds rate from 5.75 percent to
5.5 percent. The discount rate, which had
been raised to 5.25 percent at the first
meeting of the year, was left unchanged. In

BETTER CONTROL THROUGH
AN EXPLICIT OBJECTIVE
Long-term bond yields decreased as
the FOMC’s federal funds rate target
increased in late 1994 and early 1995. In
July, long-term yields rose after the fed
funds rate target was lowered. Although
politically unpopular, increases in the fed
funds rate target may be needed to lower
inflation expectations, and thus, long-term
bond yields. The focus on long-term
interest rates highlights expectations in the
monetary transmission mechanism. The
FOMC’s practice in 1995 was directed
toward the control of inflation and was
generally explained in terms of the expectations for and outcomes of the various
price indexes. Modern theories in macroeconomics and finance suggest that the
Fed could improve control over inflation
by committing to an explicit long-term
inflation (price level) objective.

REFERENCES
Bryan, Michael F., and William T. Gavin. “A Different Kind of Money
Illusion: The Case of Long and Variable Lags,” Journal of Policy
Modeling (October 1994), pp. 529–40.
Carlson, John B. “The Stability of Money Demand, Its Interest
Sensitivity, and Some Implications for Money as a Policy Guide,”
Economic Review, Federal Reserve Bank of Cleveland (1989:3),
pp. 2–13.
Friedman, Milton. “The Role of Monetary Policy,” The American
Economic Review (March 1968), pp. 1–17.
Gavin, William T., and Finn E. Kydland. “Endogenous Money Supply and
the Business Cycle,” Federal Reserve Bank of St. Louis Working Paper
95-010B (December 1995).
Hallman, Jeffrey J.; Richard D. Porter; and David H. Small. “Is the Price
Level Tied to the M2 Monetary Aggregate in the Long Run?” The
American Economic Review (September 1991), pp. 841–58.
Litterman, Robert B., and Laurence Weiss. “Money, Real Interest Rates,
and Output: A Reinterpretation of Postwar U.S. Data,” Econometrica
(January 1985), pp. 129–56.

FEDERAL RESERVE BANK

46

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

_______; David H. Small; Volker Wieland et al. “ A Quantitative
Exploration of the Opportunistic Approach to Disinflation.”
Paper presented at the Konstanz Seminar on Monetary Theory and
Monetary Policy, June 4–7, 1996.

McNees, Stephen K. “How Accurate Are Macroeconomic Forecasts?”
New England Economic Review (July/August 1988), pp. 15–36.
Moore, George R.; Richard D. Porter; and David H. Small, “Modeling the
Disaggregated Demands for M2 and M1: The U.S. Experience in the
1980s,” in Financial Sectors in Open Economies: Empirical Analysis
and Policy Issues, Peter Hooper, Karen H. Johnson Donald L. Kohn et
al., eds., Board of Governors of the Federal Reserve System, 1990,
pp. 21–105.

Pakko, Michael R. “The FOMC in 1993 and 1994: Monetary Policy in
Transition,” this Review (March/April 1995), pp. 3–25.
Sargent, Thomas J. Rational Expectations and Inflation. Harper & Row,
1986.

Orphanides, Athanasios, and David W. Wilcox. “The Opportunistic
Approach to Disinflation.” Manuscript, Board of Governors of the
Federal Reserve System, May 1996.

MEMBERS OF THE FOMC IN 1995
At any given time, the Federal Open Market Committee consists of 12 voting members. The Committee includes all seven members of the Board of Governors of the
Federal Reserve System, as well as five of the 12 presidents of the regional Federal
Reserve banks. Reflecting the importance of the Federal Reserve Bank of New York in
policy implementation, the president of that Reserve Bank is always a voting member
and is, in fact, elected as Vice Chairman of the Committee (the Chairman of the Board
of Governors is elected as Chairman of the FOMC). The remaining four positions rotate
among the presidents of the other 11 Federal Reserve banks. Although only a limited
number of Federal Reserve Bank presidents are voting members of the Committee, all
12 attend the meetings and participate in the discussions. John P. LaWare attended the
first meeting of 1995, later resigning from the Board, and was not replaced until 1996,
so there were only 11 voting members for the last seven meetings of 1995.

Listed below are the voting members of the FOMC in 1995.
Alan Greenspan, Chairman, FOMC

Lawrence B. Lindsey

Chairman, Board of Governors

Member, Board of Governors

William J. McDonough,
Vice Chairman, FOMC

Thomas C. Melzer

President, Federal Reserve Bank of
St. Louis

President, Federal Reserve Bank of
New York

Cathy E. Minehan

President, Federal Reserve Bank of Boston

Alan S. Blinder

Member, Board of Governors

Michael M. Moskow

President, Federal Reserve Bank of
Chicago

Thomas M. Hoenig

President, Federal Reserve Bank of
Kansas City

Susan M. Phillips

Member, Board of Governors

Edward W. Kelley, Jr.

Member, Board of Governors

Janet L. Yellen

Member, Board of Governors

John P. LaWare*

Member, Board of Governors
*Resigned effective April 30, 1995

FEDERAL RESERVE BANK

47

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

William R. Emmons is a research economist at the Federal Reserve Bank of St. Louis. Kent A. Koch provided research assistance.

Price Stability
and the
Efficiency of the
Retail Payments
System

of float to the user of a payment instrument—that is, the interest that can be
earned while a payment order is in the
process of clearing but has not yet been
settled—is proportional to the prevailing
interest rate as follows:

William R. Emmons

Price stability would almost certainly result in relatively low nominal short-term
interest rates because nominal interest
rates and inflation have historically moved
up and down together (see Figure 1).
Thus float benefits would be relatively
small with stable prices.
Why does float reduce the efficiency of
the payments system? A policymaker
might be tempted to ignore float because it
causes a pure transfer to occur. That is, the
interest earnings gained by one party correspond to losses of interest income by the
payment counterparty (or counterparties).
The parties to any transaction involving
float could in principle negotiate a contract that compensates the loser(s) so that
no intervention would be necessary.
The actual net result of float is far
from benign, however, because it encourages the use of float-intensive payment instruments instead of other instruments
that generate less float. Negotiation to allocate the costs and benefits of various
payment methods and instruments is difficult because there are typically more than
two parties to every transaction, involving
not just the parties making and receiving
payment, but also the primary payment
provider and potentially several other intermediaries in the payment-clearance and
payment-settlement chains. Although payment providers and other intermediaries
(primarily banks) or receivers (such as
merchants) could levy fees or penalties to
discourage the use of particular payment
instruments, this practice is not widespread. Instead, originators of payments in

User Float Benefit = (Payment Amount) ×
(Days Spent in Clearing) ×
(Daily Interest Rate on Investable Funds)

T

wo of the Federal Reserve’s most important policy mandates are to foster
price stability and to oversee the integrity and efficiency of the payments system. Effective price stability is usually
defined as annual rates of inflation or deflation close to zero. The integrity and efficiency of the payments system, on the other
hand, are related to the reliability and cost
effectiveness of the institutions and practices that individuals, businesses, and government entities use to transfer funds to one
another. Although these policy goals are seldom discussed together, this article illustrates how price stability could enhance
the efficiency of the payments system. The
argument is reminiscent of and extends
Friedman’s (1969) analysis of the optimum
quantity of money.
To simplify the discussion, I focus
only on the efficiency-enhancing effects of
price stability on the retail payments system, which consists of all those transfers
of funds that involve individuals, nonbank
firms, and government entities.1 It is quite
likely that the integrity (that is, reliability)
of the payments system would be increased as a result of the greater financial
stability that would accompany stable
prices, but I do not consider that issue
here.
The key link between price stability
and the efficiency of the retail payments
system is the level of nominal short-term
interest rates. This is because the benefit

FEDERAL RESERVE BANK

49

OF

S T. L O U I S

1

Funds transfers that occur strictly
among banks are termed wholesale payments. I do not discuss
the wholesale payments system
in this article.

S E P T E M B E R /O C T O B E R 1996

methods in use today reveals that the
most float-intensive payment methods are
also typically the most costly to produce
and process in terms of economic resources used—namely capital, labor, and
land.4 The most important examples in
the United States of float-intensive and
cost-inefficient retail payment instruments
are checks and credit cards, which are
the most frequently used retail payment
methods after cash. Thus a systematic
preference for float-intensive payment
instruments by originators of payments
and an unwillingness or inability of other
parties to discourage it reduce overall
payments-system efficiency.
The correlation between user float
benefits and real-resource costs is not
surprising when one considers that both
are a reflection of a relatively long and
often complex sequence of processing
stages, many of them labor intensive.
Checks, in particular, may require several
rounds of processing because several banks
may be involved and virtually every check
is physically returned to the bank whose
customer wrote it, wherever that may be.
Clearly, float is not the only impetus
for excessive use of costly retail payment
instruments in the United States. In addition to the forementioned fact that actual
costs are often not fully passed through to
the party controlling the choice of instrument, thus encouraging overuse, checks
and credit cards provide users with purchasing flexibility in the form of ready
access to short-term credit, as well as a
familiar and trusted system for keeping
records of transactions. These practices
and features reflect the outcome of a competitive market for financial services, so
there is little basis for direct policy intervention to change retail payment practices,
even though they do not seem to be the
most efficient use of economic resources.
Hence a monetary policy geared toward
price stability and the reduction of float
incentives to choose particular payment
instruments is a potentially important way
for the Federal Reserve to fulfill its mandate of enhancing the efficiency of the
payments system.

Figure 1

Inflation and Interest Rates, 1960-95
Percent %
16
14

Annual Inflation (C.P.I.
3-month T-Bill yield

12
10
8
6
4
2
0
1960
2

Float benefits have declined in
recent years due to generally
lower nominal interest rates
and faster collection of checks.
Recent survey evidence indicates that float still plays a role
in payment-instrument choice
for at least some consumers
and businesses, however. For
example, 8 percent of creditcard users cite “30 days to
pay” as their primary reason
for using their cards (University
of Michigan, 1995). Fifteen
percent of surveyed businesses
cited concerns about the loss of
float as the reason that their
companies had chosen not to
use financial electronic data interchange (E.D.I.) as a means
of payment, even though E.D.I.
may generate significant cost
savings in other ways (Knudson, Walton, and Young,
1994, p. 275).

3

See Humphrey and Berger
(1990), Table 2-A2, p. 82,
note f, or the Appendix to this
article, note c.

4

Humphrey and Berger (1990,
p.49) define total social or real
resource costs of payment instruments as the sum of (1)
production costs, which include
costs of fabricating and distributing the physical payment instruments, and (2) processing
costs, including the costs of accepting, safeguarding, and

1970

1980

1990

the United States generally retain some
control over the choice of payment
method and do not usually face an explicit
price penalty when choosing riskier or
costlier payment methods. Therefore, the
net user costs of alternative payment instruments, including float benefits, are potentially important in determining actual
payment practices.2
Float-influenced choice of payment instruments is inefficient for at least two
reasons. First, float is an imperfectly
priced loan (sometimes unpriced), and it
therefore increases the riskiness of the
payments system. For example, a merchant who accepts a check in payment is
actually extending credit to the customer.
The merchant’s loan is repaid not when
the merchant deposits the customer’s
check, but (typically) a few days later
when the merchant’s bank makes the
funds available for withdrawal. Bad checks
(those written against insufficient funds in
the customer’s account) represent less than
nine-tenths of one percent of all checks
written, but a typical merchant is ill-suited
to judging the credit risks presented by
any given customer.3 Thus although accurate pricing of credit is essential to prevent
concentrations of risk and potential disruptions to the payments system, checks
in fact generate a substantial volume of
mispriced credit in the form of float.
This article focuses on a second
reason why float is harmful. Available
empirical evidence on the relative cost
efficiency of the various retail payment

FEDERAL RESERVE BANK

50

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

This article begins with a description
of the current U.S. retail payments system,
paying particular attention to the float intensity and cost efficiency of the various
payment instruments (section 1). Section
2 explores the role of float in the retail
payments system. A series of examples
illustrates that price stability and low
nominal interest rates could in some circumstances improve the cost efficiency of
the U.S. retail payments system by reducing users’ incentives to choose float-intensive payment instruments. The third
section considers whether float incentives
are large enough to be considered economically significant. Section 4 concludes.

difficult to estimate. Humphrey and Berger
concluded that, for 1987, approximately
279 billion transactions, or 83.5 percent of
retail payments, were made in cash (that
is, currency or coin). Another 52 billion,
or 15.7 percent, were check or credit card
transactions.5 The remaining eight-tenths
of one percent of retail payments were
completed with a variety of instruments,
including debit cards, traveler’s checks,
money orders, automated clearinghouse
(ACH) transfers, wire transfers, point-ofsale (POS) electronic transfers, financial
electronic data interchange (EDI), electronic benefit transfers (EBTs), automated
teller machines (ATMs), home banking
by telephone or computer, Internet-based
cybercash, and stored-value cards. Wholesale (interbank) payment transactions,
which account for the bulk of the dollar
volume in the U.S. payments system, are
carried out using wire transfers, ACHs
and, to a limited extent, the exchange of
paper-based instruments such as checks
and credit card receivables.
A recent survey of consumers found a
much smaller role for cash in retail payments and correspondingly larger roles for
noncash payment instruments (University
of Michigan, 1995). The survey, conducted
on behalf of the Federal Reserve System,
found that cash accounted for only 60 percent of consumer transactions, whereas
checks accounted for 30 percent, credit
cards for 8 percent, and debit cards for 2
percent in mid-1995. All other forms of
payment accounted for less than 1 percent
of transactions.
Although retail payment practices may
indeed have changed substantially since
1987, the extent of the decline in cash’s
importance, estimated in this survey, may
be overstated. Most important, undersampling of certain groups in the population is
likely to exert a downward bias on the estimated share of cash in total transactions.
Participants in the underground economy
are likely to be underrepresented in the
sample, as are individuals who do not
maintain banking relationships. Members
of these groups rely to a great degree (or
exclusively) on cash for making payments

THE U.S. RETAIL
PAYMENTS SYSTEM
The most recent comprehensive empirical examination of the U.S. retail payments system was carried out by David B.
Humphrey and Allen N. Berger (1990),
using data for 1987. According to Humphrey
and Berger, American consumers, businesses, and government entities made about
334 billion payments for goods, services,
and financial transactions in 1987, representing a total transfer of funds of about
$342 trillion. Since gross domestic product
(GDP) in 1987 was $4.69 trillion, it follows that the annual volume of payments
was about 73 times as large as GDP. If that
same ratio prevailed in 1995, the total
value of payments in the U.S. economy
would have been $528 trillion, or about $2
million per U.S. resident. Although much
of this turnover occurs in the wholesale
payments system in association with financial markets, the basic economic fact remains that large amounts of payments are
associated with producing output and
earning income in an advanced economy.
The retail share of total payment activity—that is, all transactions except those
strictly between banks—is about 99
percent when measured by number of
transactions and about 18 percent when
measured by the dollar volume of transfers. The precise breakdown of retail
payments activity by instrument is very

FEDERAL RESERVE BANK

51

OF

S T. L O U I S

transmitting payment instruments for purposes of collecting good funds from the payor.
Both production and processing
costs ultimately entail outlays
for capital, labor, and land or
buildings.
5

See Humphrey and Berger
(1990, p. 77). Virtually all
transactions made by cash,
check, or credit card are retail
payments as defined here.
Checks and credit cards accounted for 14.2 percent and
1.5 percent of retail payments,
respectively.

S E P T E M B E R /O C T O B E R 1996

6

Humphrey and Berger (1990,
pp. 80–5. See also the Appendix.) estimated that total
(gross) float benefits were
about $55 billion in the U.S.
payments system in 1987,
whereas GDP was $4.69 trillion.
To arrive at this estimate for
float benefits, Humphrey and
Berger estimated the amount of
all payments made and the
length of time each type of payment spent in clearing, then
multiplied these amounts by the
average yield on three-month
T-bills during 1987, which was
5.775 percent.

7

See Humphrey, Pulley, and
Vesala (1996, p. 915). For
the United States in 1995,
this implies a cost of between
$147 billion and $221 billion.

8

See Humphrey and Berger
(1990, pp. 77–9, 80–5).
Also see the Appendix.

9

Payment finality refers to completion of an economic exchange with no recourse by either party, except in case of
fraud. Finality contrasts with
provisionality, as when a check
is presented in payment. When
payment is provisional, the
transaction may still be reversed within a certain period
if, for example, the checkwriter has insufficient funds in
the deposit account to cover
the payment.

10

See Humphrey and Berger
(1990, pp. 80–5). Also see
the Appendix.

11

See Humphrey and Berger
(1990, pp. 53–4, 80–5,
and 86).

12

See Humphrey and Berger
(1990, p. 54).

and are notoriously difficult to reach and
to interview. The estimates in Humphrey
and Berger may be more reliable because
they do not rely solely on consumer survey data. Instead, they construct estimates
of cash usage by combining consumer, retail, and financial-institution data. The Appendix contains details on Humphrey and
Berger’s methodology.

earned by cash’s issuer, the government) because cash balances in one’s wallet or purse
earn no interest. Milton Friedman (1969)
stressed that positive float costs borne
by users of cash drive a wedge between
individual and social optimality because individuals and firms will hold smaller cash
balances than otherwise and will expend
real resources to compensate for their inefficiently reduced holdings of money.
Despite the attractive combination of
attributes provided by cash for many retail
transactions, consumers have, over the
years, adopted alternative payment methods to varying degrees for specific purposes, such as payment by mail or over the
telephone. The most familiar substitute for
cash in retail payments in the United States
is the paper check, which represents a consumer’s IOU backed by the funds in his
or her bank deposit account. Checks are
sometimes preferred to cash because they
are more secure (requiring endorsement by
the writer and the receiver), because they
are more flexible (just as easily written for
$1 or $999.99, given sufficient deposit
funds), and because they provide a convenient written record of transactions (summarized in the monthly account statement
prepared by the consumer’s bank, as well
as in the form of the cancelled and returned checks themselves).
Checks also provide the writer a float
benefit that varies with the prevailing interest rate, the check amount, and the time
a given check spends in the mail, in the
recipient’s cash box, and in the interbank
collection and clearing process. Estimates
for 1987 indicated that consumer checks
were outstanding (that is, generating float)
for an average of three days, whereas
checks written by businesses or government entities remained outstanding for
five days. Average float benefits for all
types of checks combined were $0.83 per
check.11 Businesses captured the largest
float benefits—$35 billion in 1987—
because they tend to write larger checks
that remain outstanding for more days.
Consumer float benefits from checks were
small—$2 billion—with government float
benefits falling near the average for all

Traditional Retail Payment Methods:
Cash, Checks, and Credit Cards
Float benefits run into the tens of billions of dollars in the United States each
year and have been equivalent to about
1 percent of GDP in recent years.6 The
real-resource cost to society of making
payments is also quite large, accounting
for between 2 percent and 3 percent of
national product annually.7 The Appendix
to this article reproduces Humphrey and
Berger’s (1990) Table 2-A2 in slightly
abridged form, providing details concerning the resource costs and float benefits associated with various payment instruments.
The three most commonly used retail
payment instruments—cash, checks, and
credit cards—generated essentially all of the
float and consumed 93 percent of all the
real resources devoted to making payments
in the United States in 1987. These same
instruments accounted for 99 percent of
transactions and 18 percent of the funds
transferred in that year.8 Thus efforts to reduce the market share of these traditional
instruments, to increase the efficiency of
these instruments, or both—especially with
regard to checks and credit cards—could
produce significant economic benefits.
Cash is currently the predominant and
most efficient instrument for making payments in the vast majority of face-to-face retail transactions. Cash provides simplicity,
anonymity, and finality to both buyer and
seller at very low cost.9 The average social
cost of using cash in retail payments
amounted to approximately $0.04 per transaction as of 1987.10 In addition to the risk of
loss or theft, one major disadvantage of cash
from the user’s point of view is that it entails
a float cost (corresponding to interest

FEDERAL RESERVE BANK

52

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

checks, about $2 billion.12 Checks and
their associated processing infrastructure
are quite costly, consuming approximately
$0.79 of economic resources for every
check written in 1987.13
Another substitute for cash (and
checks) is the credit card. Bank-issued
credit cards are sometimes co-branded, an
arrangement that allows nonbank firms to
participate in the marketing of the card and
to deliver special benefits or promotions to
cardholders. Credit cards, like checks, provide the accepting merchant with an IOU
that can be redeemed for cash or deposits
at the merchant’s bank. In practice, credit
cards provide consumers with most of the
advantages of checks, along with greater
on-the-spot purchasing power. This advantage derives primarily from the assumption
of nonpayment risk by the card-issuing
bank. The acceptance of liability differs
from the case of a check, in which the merchant bears the risk that the check cannot
be redeemed for cash or deposits. Banks are
willing to absolve merchants of this risk
when credit cards are used because banks
specialize in screening and monitoring retail customers’ creditworthiness and because merchants effectively pay a fee for
each credit card transaction they accept.14
The risks to banks involved in credit
card transactions are minimized by screening the creditworthiness of consumers both
before they receive cards and at the time of
a purchase. Credit checks before a card is
issued allow issuing banks to exclude
likely defaulters from using their card at
all. On-line data bases maintained by the
major credit card associations allow merchants to obtain an update on a consumer’s
credit status at the time of purchase. Merchants receive an authorization number
from the issuing bank at the time of purchase to verify the credit check and thereby
transfer nonpayment risk to the bank.
Credit cards provide a significant
amount of float if they are used exclusively
as payment instruments. This is true of socalled convenience users of credit cards,
those cardholders who pay off all charges
within the grace period granted by the
issuer. Use of the card as an instrument

for taking out preapproved consumer
loans, on the other hand (that is, running
a balance), redirects the float benefit from
the consumer to the issuing bank. Float
benefits averaged about $0.44 per credit
card transaction in 1987, considerably
more than the estimated $0.07 of float
benefit per check written by consumers
or the $0.05 of float cost per cash transaction.15 Credit card payments are, like
checks, relatively costly to process because they require a complex supporting
infrastructure and multiple rounds of
processing by the parties seeking to collect the ultimate payment in bank funds.
Humphrey and Berger estimate that the
average credit card payment consumes
about $0.88 in economic resources.16

Non-traditional Retail Payment
Methods: ACH, Debit Cards, and
Stored-Value Cards
Although the majority of U.S. consumers and businesses express satisfaction
with their traditional payment options
(cash, checks, and credit cards), significant
efforts are currently being made by many
nonfinancial firms, banks, trade associations, and government agencies such as the
Federal Reserve to expand the use of newer
retail payment methods. In many instances,
the technology underlying the newer payment methods has existed for some time,
although it has not been in common use.
In other cases, advances in information
technology and communications have accelerated the development of these substitutes for cash, checks, and credit cards.
New methods of making consumer
payments include ACHs, debit cards, ATM
cards, stored-value cards, and cybercash,
a payment instrument that exists on,
and is used solely to make purchases over,
computer networks. New methods for
making business and government payments include ACH, financial EDI,
purchasing cards, and EBT.
Among the most important of the
newer retail-payment methods is the use of
automated clearinghouses. ACH transfers
are used to effect direct deposit of pay-

FEDERAL RESERVE BANK

53

OF

S T. L O U I S

13

See Humphrey and Berger
(1990, pp. 80–5). See the
Appendix.

14

The credit card fee is collected
by the merchant’s bank by discounting the merchant’s credit
card receipts. For example, the
merchant trades a credit card
charge receipt of $100 for a
$97 credit in its account at the
bank. The merchant’s bank
then collects the receivable
from the card-issuing bank
through established interbank
clearing channels, receiving perhaps $98 for the item. Finally,
the issuing bank bills the cardholder periodically, for payment
of the full $100 and the accumulated balance of other
charges.

15

See Humphrey and Berger
(1990, pp. 50, 54, and
80–5). See the Appendix.

16

See Humphrey and Berger
(1990, pp. 80–5). See the
Appendix.

S E P T E M B E R /O C T O B E R 1996

17

See Bauer and Ferrier (1996,
p. 1025).

18

This statement applies strictly
to debit card networks that operate on-line, with merchants
maintaining real-time electronic
connections with banks so that
funds can be transferred as purchases are made. There are
also some debit card networks
that operate off-line, with merchants cumulating a record of
transactions during the day,
which are then transmitted at
the end of the day in a batch
to the merchant’s bank for processing and collection, perhaps
through ACH. In this case, the
consumer enjoys some float,
although the amount is generally less than with a check or
credit-card purchase.

19

See Humphrey and Berger
(1990). Recall that the estimated average social cost
per transaction using cash
was $0.04; for ACH $0.29;
for checks $0.79; and for
credit cards $0.88. See the
Appendix.

checks, Social Security checks, or other
benefits into recipients’ bank accounts; to
carry out routine repetitive payments by
consumers, such as mortgage payments,
insurance premiums or utility bills; or to
execute cash-management operations for
far-flung businesses. Tiny amounts of float
are created in ACH debit transfers (requests for payment), whereas ACH credit
transfers (sending payments) create no
float. Funds transfers made by ACH are
currently very efficient because they exploit economies of scale in information
processing. A large employer can pay its
employees, for example, simply by encoding all of the pertinent payroll information
(including the employee’s name, the amount
of the payment, and the employee’s bank
account number) and sending this information to the ACH through its bank (typically on magnetic tape or through EDI).
The funds transfers from the employer’s
account to those of all its employees are
then carried out electronically among
banks at a precisely specified time. Humphrey and Berger (1990) estimate that the average social cost of an ACH transfer was
only $0.29 as of 1987. As more and more
transactions occur through the ACH, the
cost per transaction falls, an indication that
unexploited economies of scale remain in
ACH payments.17
Another fast-growing form of retail payment is the debit card, sometimes also
called a POS card. To consumers, a debit
card functions much like a credit card at the
point of purchase, with one important exception. Rather than providing the consumer with float, as with convenience use of
a credit card, a debit card transaction transfers funds directly from the consumer’s bank
account to the retailer’s bank account. Hence
consumers receive no float benefits.18 The
lack of float associated with debit cards is
almost certainly one of the reasons why
consumers have not adopted debit cards
even more rapidly than they have. Debit
card payment systems are developing
rapidly, so it is difficult to know how costefficient they are currently. Estimates from
the late 1980s put debit card transaction
costs at about $0.47 per transaction, consid-

erably less than the transaction costs of
credit cards and checks, but greater than
those for cash and ACH.19
A new form of retail payment that has
captured the attention of many merchants,
bankers, and computer experts is the
stored-value card, also referred to as the
electronic purse or electronic cash. Storedvalue cards promise greater convenience in
certain retail-purchase situations, including
those in which coins are normally used. For
example, stored-value cards would be quite
convenient for making purchases from
vending machines, at newspaper kiosks, or
in laundromats. Stored-value cards can be
thought of as small-dollar traveler’s checks.
The consumer purchases a card from a
financial institution with monetary value
encoded on the card and then uses it anywhere merchants are willing to accept its
stored value in payment. The merchant, in
turn, trades the stored value for monetary
value in the merchant’s bank account. Finally, the bank must collect funds from the
card-issuing institution. Stored-value cards
will likely suffer in consumers’ eyes because
they incur a float cost: Consumers must
pay out funds in advance of any retail purchases, so the funds do not earn interest.
Given the recent arrival of stored-value
cards on the retail payment scene, it is impossible to gauge their cost effectiveness
with any certainty. One relevant comparison might be the resource costs associated
with traveler’s checks, which were estimated by Humphrey and Berger (1990) to
be about $1.18 per transaction. The purely
electronic nature of stored-value cards
should reduce the per-unit costs to a great
extent relative to traveler’s checks, but the
necessary investments by merchants,
banks, and consumers to support the cards
means that the economics of stored-value
cards will not be favorable until a substantial volume of payment activity is ensured.
In sum, there has been moderate
progress in converting U.S. retail payments
from a paper-based system to a more predominantly electronic system. The Appendix summarizes the resource costs and
user float benefits associated with the
major retail payment instruments as of

FEDERAL RESERVE BANK

54

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Table 1

Use of Non-Cash Retail Payment Instruments (in Billions*)

Checks
Credit Cards
ACH
Debit Cards
Total (including instruments not shown)

1988

1989

1990

1991

1992

1993

1994

50.99
8.81
1.01
0.17
61.08

52.90
8.90
1.18
0.22
63.30

55.44
10.75
1.43
0.28
67.96

57.47
11.24
1.63
0.30
70.75

58.40
11.70
1.84
0.51
72.56

60.30
12.52
2.09
N.A.
75.01

61.67
13.68
2.37
1.05
78.89

* From: Bank for International Settlements (1993, p. 472; 1994, p. 110; and 1995, p. 110)

1987. Cash, check, and credit card processing have become more automated since
1987, thereby reducing unit costs associated with these instruments. At the same
time, newer payment instruments have
achieved ever-larger transaction volumes,
reducing their unit costs.
Purely electronic payment methods,
such as debit cards, ACHs, cybercash, and
home banking, are among the fastest
growing sectors of the payments market
and may have accounted for well over
3 billion transactions in 1995.20 Table 1
shows the relative importance of several
leading noncash retail payment instruments in recent years. Figure 2 illustrates
the relatively high growth rates of debit
cards and ACH in particular.
Despite rapid growth of nontraditional
retail payment instruments, it is still much
too early to describe the changes in U.S.
consumers’, businesses’, and government
entities’ payment habits to date as a true
revolution. After all, it is likely that 60 percent to 80 percent of retail payments are
still made in cash and that 15 percent to
30 percent are made with checks or credit
cards.21 All other payment instruments together probably account for 10 percent of
all transactions or less. Hence the retail
payments revolution has just begun.

Figure 2

Total Growth in Use of Non-Cash Payment
Instruments 1988-1994
Percent (%)
600
500
400
300
200
100
0

Debit cards

ACH

Credit cards

Checks

effective price stability could make the retail payments system more efficient. The
reason is that, with no sustained inflation
and thus consistently low short-term
nominal interest rates (averaging perhaps
2 percent to 3 percent), the benefits of
float that create incentives for payment
originators to choose float-intensive instruments would be much less than they
are today. Modest inducements from retailers and payment providers to use the
more efficient retail payment instruments,
such as merchant surcharges on checks
and credit cards or discounts for cash and
debit cards, might then be effective in
changing payment practices. Although
payment-instrument choice would be
most efficient if each instrument’s actual
resource cost were passed through to the
party choosing the instrument—typically
the party making payment—even partial
pass-through of costs would be more effective in an environment of price stability. Thus this article extends Friedman’s
(1969) analysis of the welfare costs of in-

THE ROLE OF FLOAT ON
CONSUMER INCENTIVES
IN RETAIL PAYMENTS
This article’s theme is that a movement away from today’s environment of
moderate inflation to one characterized by

FEDERAL RESERVE BANK

55

OF

S T. L O U I S

All instruments

20

See Bank for International
Settlements (1993, p. 472;
1994, p. 110; 1995, p. 110).

21

See note 20, Humphrey and
Berger (1990, pp. 77–9),
and University of Michigan
(1995).

S E P T E M B E R /O C T O B E R 1996

potential of different levels of nominal interest rates and float benefits to affect decision makers’ behavior when they bear
some or all of the social costs resulting
from their decisions.

Table 2

Payment
Instrument
Cash
Check

Float Benefit at22

Days in
Clearing

5 Percent

2 Percent

0
3

0
$0.01

0
$0.004

Example 1: Small-Value Personal Payment
Transaction. A consumer purchases a

restaurant meal for $25. She may pay
with cash or check. Assume first that the
only relevant cost to the customer is float;
in other words, there are no per-check
fees and the customer has at least $25
in her wallet (hence no need to pay any
ATM withdrawal fees). The customer
receives no float if she pays with cash,
whereas she expects three days of float if
paying by check. Which retail payment
instrument would she choose if the shortterm interest rate were 5 percent? What
if it were 2 percent?
Float benefits associated with each instrument are shown in Table 2. The rational decision maker prefers to write a check
whatever the interest rate, although the
float benefit is trivial.
Adjusting Humphrey and Berger’s
(1990) estimates for 1987 to account for
general inflation through 1995, social costs
associated with each of the retail payment
instruments considered here are shown in
Table 3.23 As the table indicates, the customer’s choice of a check, though modestly
beneficial to herself, is quite costly from the
standpoint of overall efficiency of the payments system. In this case, the consumer
might switch her payment preference to
cash in response to even a small incentive,
such as a 1 percent cash discount offered by
the restaurant (that is, 25 cents off the bill).
In this example, the potential interest
earnings created by float would obviously
play a very minor role in a typical decision
maker’s choice of payment instrument. In
fact, this is true of most consumer payments. A more important reason why a
consumer might use a check instead of
cash is the flexibility it affords. Suppose,
for example, that the restaurant diner had
no cash in her pocket and no money in
her bank deposit account. Suppose further that the customer knew that her pay-

Table 3

Payment Instrument

Average Social Cost
(in 1995 Prices)

Cash
Check

$0.05
$1.06

Table 4

Float Benefit at24

Payment
Instrument

Days in
Clearing

5 Percent

Debit Card
Credit Card

0
15

0
$1.23

2 Percent
0
$0.49

Table 5

Payment Instrument
22

Float benefit is the amount of
the payment, $25, multiplied by
the number of days the payment
takes to clear, multiplied by the
daily interest rates—5 percent
or 2 percent divided by 365.

23

These are average rather than
marginal social costs, which
one would normally prefer
when analyzing a given consumer payment decision. The
illustrations here are intended
to point out the full-cost implications of various payment
practices over time, however,
so average costs are the relevant measure.

24

Float benefit is the amount of
the payment, $600, multiplied
by the number of days the payment takes to clear, multiplied
by the daily interest rates—5
percent or 2 percent divided by
365.

Debit Card
Credit Card

Average Social Cost
(in 1995 Prices)
$0.63
$1.18

flation to the full range of retail payment
instruments in use today.
To illustrate the role of float and the
potential importance of price stability
in the retail payments system, I consider
three examples. The examples are designed
to capture the roles played by float and explicit transactions costs in a small-value
personal payment transaction, in a largevalue personal transaction, and in a largevalue business transaction, respectively, in
determining which payment instrument a
rational decision maker would be likely to
choose. First, I determine the incentives the
decision maker would face as a result of
float. Then I compare the privately optimal
choice of payment instrument with the
most efficient choice from the standpoint
of total social costs. Finally, I examine the

FEDERAL RESERVE BANK

56

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Table 6

Float Benefit at

Payment
Instrument

5 Percent

Debit Card
Credit Card

0
$1.23

Total User Benefit at25

2 Percent

Average Social Cost and
User Price (1995 Prices)

5 Percent

2 Percent

0
$0.49

$0.63
$1.18

2$0.63
$0.05

2$0.63
2$0.69

check would be deposited in her bank account the next day through ACH. Being
able to write a check that will not clear for
three days can thus be very valuable in
some situations. Taking into account the
value of flexibility, one can see that the
short-term rate may actually understate
the true economic value of float.

cially efficient choice, which would be to
use a debit card.
Would full pass-through of the average
social cost of the payment instrument
change this rational consumer’s choice of
instrument? It turns out that the answer
depends on how much float is available.
In an environment such as the one that
exists today with a moderate level of inflation and short-term interest rates around
5 percent, the consumer’s choice would be
unaffected. The situation would be quite
different in an environment of price stability and a low interest rate of 2 percent,
however. The calculations that a rational
decision maker would make if he were
forced to pay the full social cost of the payment instrument are shown in Table 6.
One conclusion from the table is striking: Even if a rational consumer were forced to
bear the full cost of his choice of payment instrument, he would still choose a credit card if
the nominal interest rate were 5 percent, but
he would choose a debit card if the interest
rate were 2 percent. This example points out
that, for at least some retail payment transactions, even the implementation of full passthrough of costs to the purchaser would
not eliminate incentives for inefficient payment-instrument choice when float benefits
remain substantial. Even with only partial
pass-through of costs to payment users,
lower interest rates can shift a consumer’s
incentives in the direction of more efficient
payment instruments. The next example
illustrates a clear-cut case in which price
stability would be likely to have a big effect.

Example 2: Large-Value Personal Payment
Transaction. Suppose a new homeowner

purchases a home appliance, such as a
refrigerator, for $600. The consumer
may pay with a debit to his deposit account using a debit card or with a credit
card, both of which the appliance store
accepts without surcharge or discount.
Assume first that the only relevant cost
to the purchaser is float. In other words,
there are no transaction fees for either
type of payment. The customer receives
no float if he pays with his debit card,
though he expects 15 days of float if he
pays with a credit card. Which payment
instrument would a rational decision
maker choose if the short-term interest rate
were 5 percent? What if it were 2 percent?
Float benefits associated with each instrument are illustrated in Table 4. In this
case, it is likely that the consumer would be
aware of the benefit of using a credit card to
capture float and that he might do so.
Is the customer’s choice of payment
instrument efficient from the standpoint
of the economy’s use of resources? Again
adjusting Humphrey and Berger’s (1990)
estimates for 1987 to account for general
inflation through 1995, Table 5 shows average social costs associated with the use
of debit and credit cards. Comparing the
user float benefits to average social costs,
one can see that the consumer’s personal
preferences are again the reverse of the so-

Example 3: Large-Value Business Payment
Transaction. Suppose a small-business owner

is preparing the firm’s monthly payroll covering 10 employees with an average wage payment of $2,000 each (that is, a total payroll of

FEDERAL RESERVE BANK

57

OF

S T. L O U I S

25

Total user benefit is the user’s
float benefit at the given interest rate minus the social cost
of each payment, which is also
the user’s price.

S E P T E M B E R /O C T O B E R 1996

rate, the less valuable is a float-intensive
payment instrument to the user. With little
float benefit available from any payment instrument, originators of payments might
shift their payment preferences away from
float-intensive instruments, given even a
modest inducement to do so. Because the
per-item social costs are roughly the same
no matter how large the dollar amount of
the payment involved, this general benefit
arising from price stability would be particularly important in changing payment patterns for items involving only a few dollars.

Table 7

Payment
Instrument
ACH
Check

Days in
Clearing
0
3

Float Benefit26
at 5 Percent at 2 Percent
0
$0.82

0
$0.32

Table 8

Payment Instrument
ACH
Check

26

Float benefit is the amount of
the payment, $2,000, multiplied by the number of days
the payment takes to clear,
multiplied by the daily interest
rates—5 percent or 2 percent
divided by 365.

27

Humphrey and Berger (1990,
pp. 54, 77, 80). This figure is
the sum of social costs accounted for by all checks and
credit-card transactions
($41.85 billion) less the total
cost of making 30.9 billion
payments with debit cards at
47 cents per transaction
($14.52 billion) and 21.2 billion payments via ACH at 29
cents each ($6.15 billion). If
some check and credit-card
transactions were instead replaced by cash payments, the
cost savings would be even
greater.

Average Social Cost
(in 1995 Prices)
$0.39
$1.06

ARE FLOAT INCENTIVES
ECONOMICALLY
SIGNIFICANT?

$20,000). The owner may pay her employees
using ACH or checks. First, the owner examines the float implications of the two payment methods. She receives no float if she
pays by ACH, whereas she expects three days
of float on average if she pays by check.
Which method would this rational business
owner choose if the short-term interest rate
were 5 percent? What if it were 2 percent?
Float opportunity costs per employee
are shown in Table 7. Now consider the
bank fees associated with each instrument.
Assume that the pricing schedule from the
owner’s bank reflects average social costs
exactly. Adjusting Humphrey and Berger’s
(1990) estimates for 1987 to account for
general inflation through 1995, the total
bank fees assessed for using each method
would be as shown in Table 8. The smallbusiness owner’s total costs are therefore
shown in Table 9. The net cost of paying
her employees is $0.15 less per person by
check than by ACH if the interest rate is 5
percent, though the small-business owner
saves $0.35 per employee by using ACH
under the low–interest rate scenario. This
example illustrates that, even if full social
costs are passed through to the decision
maker, moderate levels of inflation and interest rates—such as we currently have in
the United States—may be sufficient to
distort incentives toward socially wasteful
payment instruments.
To summarize the point made in these
examples, the lower the short-term interest

How important is the choice of retail
payment instrument? Consider the following thought experiment: If all 1987
household payments by check (totalling
25.8 billion) and by credit card (5.1 billion) had instead been carried out by debit
card, and if all business and government
check payments (21.2 billion) in that year
had been executed via ACH, those who
made the payments would have lost about
$41 billion in float benefits. These “losses”
would have been exactly balanced by $41
billion of reduced float costs incurred by
payment recipients, however, so there
would have been no net gain or loss for
society from this source. At the same time,
the total resource cost of making retail
payments in the United States would have
been about $21 billion lower.27
This estimate must be interpreted as
an upper bound, since some of the social
costs allocated to each payment instrument were already sunk as of 1987, and
therefore would have been incurred even if
the payment instruments weren’t used.
The best interpretation of this $21 billion
estimate is that it represents the efficiency
gain that might have resulted if the retail
payments system had developed differently
than it actually did: namely, without
checks and credit cards being used in retail payments at all.
Could it be that checks and credit
cards appear to be “resource guzzlers” but

FEDERAL RESERVE BANK

58

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Table 9

Payment
Instrument
ACH
Check

Float Benefit at
5 Percent

2 Percent

Average Social Cost and
User Prices (1995 Prices)

0
$0.82

0
$0.32

$0.39
$1.06

Total User Benefit at28
5 Percent

2 Percent

2$0.39
2$0.24

2$0.39
2$0.74

tail payments system on the basis of this
(or any other) individual study, however.
First, the Wells (1996) and Humphrey and
Berger (1990) studies are not directly comparable, so conclusions drawn from one
study cannot be overturned by the other.
For example, Wells uses different data
sources to calculate some key components
of social costs associated with checks and
ACH (the only two payment methods examined in her study). The fact that Wells
finds much higher social costs than do
Humphrey and Berger may be attributable
to underlying trends, or it may be due
simply to differences in data-collection
methods. Similarly, Wells finds much
lower float benefits than do Humphrey
and Berger, but the former study assumes
a short-term interest rate of 3 percent (as
it was during 1993), while the latter study
assumes a rate of 5.775 percent, which is
much closer to the average rate that has
prevailed over the last ten years (and
which exists today).
Even more importantly, Wells uses different definitions of social costs and float
benefits than do Humphrey and Berger.
For example, Wells includes payee costs
(representing over 40 percent of her estimate of the social cost of checks), while
Humphrey and Berger exclude them altogether. Wells excludes mail float and recipient float, while Humphrey and Berger
include both (representing 37 percent and
24 percent of their estimate of float benefit, respectively). Finally, Wells assumes all
checks require only one round of clearing
through a financial institution, while
Humphrey and Berger allow for two stages
(the second stage representing 20 percent
of their estimate of float benefit). It is
important to note that each of these
methodological differences has the effect

in fact deliver correspondingly higher satisfaction to their users and recipients? In
other words, could they be the “Cadillac
of payment instruments,” simply items on
an economically efficient menu of options? If the genuine economic benefits
provided by float-creating retail payment
instruments exceed the social costs even
when the zero-sum nature of float is netted out, then there is really no efficiency
problem in the U.S. retail payments system.
A recent study concludes that checks
may be popular in the U.S. not because
they provide significant float benefits to
check-writers but, rather, because they
provide a uniquely desirable bundle of
payment services not available from other
payment instruments (Wells, 1996). This
conclusion is based on new evidence from
1993 suggesting that checks are, in fact,
even more costly than Humphrey and
Berger (1990) had estimated, while float
benefits from checks may have been virtually eliminated.29 Wells reasons that tiny
float benefits could not possibly explain
why 60 billion checks are written each
year in the United States. Hence, Humphrey and Berger’s emphasis on float as an
incentive to use checks is “a mistaken
view” (Wells, 1996, p. 3).
The implication of Wells’ conclusion,
if correct, is that float-intensive payment
instruments such as checks may not be
“over-used” in the U.S. at all. The higher
resource costs associated with checks and
other float-creating instruments such as
credit cards should not be labelled wasteful, since genuine value is being delivered.
In short, float does not reduce economic
efficiency, because float is essentially irrelevant for check use.
It is probably unwise to become complacent about the efficiency of the U.S. re-

FEDERAL RESERVE BANK

59

OF

S T. L O U I S

28

Total user benefit is the float
benefit at the given interest rate
minus bank fees.

29

Wells (1996, p. 5) estimates
that the average social cost of a
payment by check in 1993 was
$2.93 (midpoint of estimated
range) while Humphrey and
Berger (1990, p. 80, and Appendix of this article) conclude
that a check payment cost society $0.79 in 1987 (or $1.00
in 1993 dollars). Wells estimates 1993 float benefits for
checkwriters of only $0.09, versus Humphrey and Berger’s estimate of $0.83 ($1.04 in
1993 dollars).

S E P T E M B E R /O C T O B E R 1996

30

In other words, efficient instrument choice by all consumers
and businesses is assumed
rather than being demonstrated. Wells (1996, p. 4)
appeals to the Coase Theorem
and relatively inexpensive
bargaining among concerned
parties.

of increasing Wells’ estimate of social costs
or decreasing her estimate of the float benefits of checks. Thus, even if both studies
used identical data, Wells would find a significantly higher social cost and lower
float benefit for checks than do Humphrey
and Berger.
Apart from these differences in data
collection and methodology that prohibit
direct comparisons between the Wells and
Humphrey and Berger studies, there are
more obvious reasons to doubt that float
incentives are irrelevant for retail payment-instrument choice. As noted above,
at least some consumers and businesses
are aware that checks and credit cards create float, even if they do not calculate the
(often trivial) benefit float confers. Surveys
indicate that float incentives can and do
influence payment-instrument choice for
some payors at least some of the time
(University of Michigan, 1995; Knudson,
Walton, and Young, 1994).
The Wells study assumes that total social costs are fully internalized (taken into
account) by all the relevant parties in a retail payment decision. Consequently, actual instrument choices reflect genuine
benefits received.30 In practice, of course,
negotiation over the form of payment is
the exception, rather than the rule. Repeated negotiation among all involved
parties in the retail payment clearing and
settlement chain would be extremely
costly. Thus, in general, the payor exercises some choice of instrument from
among a menu of alternatives provided by
the payee. Since costs are incurred by others in processing the chosen payment instrument, the decision maker does not fully
internalize the social costs of the payment
instrument. In sum, it is not appropriate to
assume, as Wells does, that all social costs
associated with a particular payment instrument are fully considered—and matched
by real benefits—in the choice of instrument.
Finally, the fact that consumers and
businesses like certain features of a given
payment instrument does not imply that
all facets of that instrument contribute to
(or are neutral with respect to) social wel-

fare. For example, surveys reveal that consumers and businesses clearly appreciate
certain features of checks and credit cards,
such as convenience in record-keeping and
ease of use in making payments through
the mail (checks) or over the phone
(credit cards). However, there is no inherent reason why float must be associated
with payment instruments that provide
these or other features. Humphrey, Pulley,
and Vesala (1996. p. 926) point out that
most retail payments in European countries and Japan are made without creating
float for the payor. This does not prevent
consumers and businesses in those countries from keeping records or making payments through various communication
channels.

CONCLUSION
Price stability and the efficiency of
the payments system are two of the Federal Reserve’s most important policy goals.
This article suggests that price stability
could enhance the efficiency of the U.S.
retail payments system because lower
nominal interest rates would reduce float
benefits, which affect payment users’ incentives to choose particular payment
instruments. Friedman (1969) pointed
out the real-resource costs incurred by
economic agents holding inefficiently
small real money balances, by which he
apparently meant cash. This article extends the argument to the full range of
retail payment instruments. Instrument
choice matters for economic efficiency
because the most float-intensive payment
instruments are typically also the most
costly to produce and process, reflecting
relatively long, labor-intensive clearing
and settlement cycles.
Incomplete pass-through of retail payment costs to the decision makers who
control instrument choice is a widespread
practice that may well continue. Because
this practice is the outcome of a competitive retail payments marketplace, there is
little that policymakers can or should do
to change it. A monetary policy geared toward price stability may therefore be an

FEDERAL RESERVE BANK

60

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

important practical way in which the Federal Reserve can make progress toward its
goal of increasing the efficiency of the payments system.

REFERENCES
Bank for International Settlements. Payment Systems in the Group of
Ten Countries, 1993.
_______. Statistics on Payment Systems in the Group of Ten Countries, 1994.
_______. Statistics on Payment Systems in the Group of Ten Countries, 1995.
Bauer, Paul W., and Gary D. Ferrier. “Scale Economies, Cost Efficiencies,
and Technological Change in Federal Reserve Payments Processing.”
Journal of Money, Credit, and Banking (November 1996 Part 2), pp.
1004-1039.
Friedman, Milton. The Optimum Quantity of Money and Other Essays.
Aldine Publishing Company, 1969.
Humphrey, David B., Lawrence B. Pulley, and Jukka M. Vesala. “Cash,
Paper, and Electronic Payments: A Cross-Country Analysis,” Journal of
Money, Credit, and Banking (November 1996 Part 2), pp. 914-39.
_______ and Allen N. Berger, 1990. “Market Failure and Resource
Use: Economic Incentives to Use Different Payment Instruments,” in
David B. Humphrey, Ed., The U.S. Payment System: Efficiency, Risk
and the Role of the Federal Reserve (Boston: Kluwer Academic Publishers), 45-86.
Knudson, Scott E., Jack K. Walton II, and Florence M. Young. “Businessto-Business Payments and the Role of Financial Electronic Data Interchange,” Federal Reserve Bulletin 80 (April 1994), pp. 269-78.
University of Michigan. “May 1995 Survey of Consumers.” Commissioned by the Federal Reserve System Future Uses of Currency Task
Force.
Wells, Kirstin. “Are Checks Overused?” Quarterly Review, Federal Reserve Bank of Minneapolis (Fall 1996), pp. 2-12.

FEDERAL RESERVE BANK

61

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Appendix

coin was assumed to be replaced every 15
years. To these replacement costs are
added the cost of producing new currency
and coin because of yearly growth in demand, a function of inflation and population growth. The yearly growth of cash
was 8 percent between 1985 and 1987. In
sum, the yearly production cost of cash is
$419 million {$306 million/2 + $1,648
million/15 + [($306 million + $1,648 million) × 0.08] = $419 million a year}.
In accepting cash for retail sales, the
payee may incur costs for all of the following: POS and accounting, theft and loss of
cash, safekeeping and security, and deposit
charges and fees paid to financial institutions. (See Curtin, Richard T. Payment
Method Costs: Assessments by Retailers. Survey Research Center, University of Michigan, 1983). These costs, expressed as a
percentage of the average cash retail transaction amount, were 2 percent (and 2.5
percent for checks). Studies have suggested that the share of retail sales paid for
with cash is around 30 percent. Applying
this 30 percent share and the 2 percent
cost percentage to all retail sales in 1987 of
$1,505 billion yields a total payee cost of
cash of $9,030 million (0.02 × 0.30 ×
$1,505 billion). This payee cost estimate
covers payee bank cash costs as well.
Payor costs of using cash cannot be reliably estimated but would include the costs
of lost cash and theft along with the cost
of obtaining cash from a bank. The bank
costs of giving cash over the counter when
checks are cashed are estimated at $1,674
million. This estimate is derived primarily
from the 1986 Functional Cost Analysis
data for the 76 banks with over $200 million in deposits. (See Board of Governors
of the Federal Reserve System. Functional
Cost Analysis, 1986). The smaller banks
were discarded to offset the bias toward
smaller banks in the Functional Cost
Analysis (FCA) sample, although the difference would be negligible. The per-transaction cost of giving cash is estimated at
$0.314. The number of checks cashed per

Resource Costs and User Float Benefits of
Retail Payment Instruments, 1987*
Type of
Payment
Instrument

Total Social or
Real Resource
Cost per Transaction

Cash
Checks
Credit cards
ACH
POS (Debit Cards)
Traveler’s Checks
Money Orders
Wire transfers

User Float Benefit
(+ for Benefit,
– for Cost)

$0.04a
0.79c
0.88e
0.29g
0.47i
1.18k
1.79m
7.33o

2$0.05b
0.83d
0.44f
0.00h
20.00j
20.00l
20.00n
0.02p

* Slightly abridged version of Table 2-A2 in Humphrey and Berger, 1990, pp. 80-5.

THE FOLLOWING IS A
DETAILED EXPLANATION
OF THE ITEMS IN THE
TABLE ABOVE:
a

Cash

Production costs were estimated by
multiplying the number of currency notes
outstanding (11,776 million) times an average production cost of $26 per 1,000
notes produced at the Bureau of Engraving
and Printing (giving $306 million). The
weighted average cost of coin issue is
$0.0107 per coin at the U.S. Mint for approximately 154 billion coins outstanding
(giving $1,648 million). This production
cost for all currency and coin outstanding
is transformed into a yearly cost as follows: Since the average $1 bill (37 percent
of currency notes outstanding) is replaced
every 1.5 years, with other denominations
replaced at more infrequent intervals, all
currency was assumed to be replaced every
two years. Coin has a very long lifetime
but requires replacement over time as individual coins are lost. For our purposes,

FEDERAL RESERVE BANK

62

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

personal account was estimated by taking
the total on-us plus transit checks cashed
by the average bank per year and dividing
by the average number of personal accounts per bank (for those banks reporting
a separate breakout by type of account),
giving 48.55 cashings per account per
year. The aggregate number of personal accounts was estimated at 109.8 million [See
Avery, Robert, Gregory Elliehausen, Arthur
Kennickell, and Paul Spindt. “Changes in
the Use of Transaction Accounts and Cash
from 1984 to 1986,” Federal Reserve Bulletin (March 1987), pp. 179–96. This article
was a special unpublished computer run.]
Total check-cashing costs were therefore
estimated at $0.314 × 48.55 × 109.8 million = $1,674 million. Federal Reserve
costs in transporting and processing currency and coin, including the cost of retiring old and counterfeit currency, were
$154 million [See Board of Governors of
the Federal Reserve System. Planning and
Control System Expense Report (PACS),
1987.] Thus the payee ($9,030 million),
payor and bank ($1,674 million), and Federal Reserve ($154 million) processing
cost of cash is $10,858 million.
This figure excludes the government
production cost of $364 million and the
portion of processing costs borne by the
Federal Reserve, $154 million, both of
which are provided free. The remaining
private-sector costs are assumed passed on
to cash customers through higher prices.
b

that the seigniorage benefits to the government were reduced and that vault cash
costs are now part of reserve requirements.

Checks

c

Check production costs were estimated
by taking the actual production costs per
standard consumer-type check ($0.025
cents per item) and business-type check
($0.05 cents per item) for a large East
Coast check-printing firm and multiplying
these average production costs by the volume of the types of checks written. The
numbers of consumer, business, and government checks to be written are estimated,
respectively, at 25.8 billion, 18.8 billion,
and 2.4 billion. [See Humphrey and Berger
(1990), Table 2-2, pp. 54–5.] Government
checks are more like business checks and
are included there. Thus the total check
production cost is estimated at ($0.025 ×
25.8 billion consumer checks) + [$0.05 ×
(18.8 billion business checks + 2.4 billion
government checks)] = $1,705 million.
Check-processing costs comprised accounting and disbursement costs of business and government payors, postage
($0.22) and envelope costs ($0.02) for all
payors (business, government, and consumers), and bank costs. The opportunity
cost of consumer payers’ time to write and
mail checks was not included because few
consumer payors actually have the opportunity of getting paid for the time saved if
they do not write checks. Accounting and
disbursing costs for business and government payors is estimated at $0.24 per
check or payment transaction, based on
the $0.239 per payment transaction cost
for the U.S. Treasury’s direct deposit program. (See Dudley, William C. A Comparison of Direct Deposit and Check Payment
Costs, ed. 2, Board of Governors of the
Federal Reserve System, 1983.) This excludes all postage, commercial bank, and
Federal Reserve check-processing and collection expenses. Thus $0.24 multiplied by
the sum of business and government
checks [21.2 billion items, according to
Humphrey and Berger (1990) Table 2-2,
pp. 54–5] gives $5,088 million in business

Cash

The opportunity cost of holding idle
coin and currency was derived from evaluating the $230 billion in coin and currency
in circulation in 1987, which is less than
the $271 billion outstanding, at the average
90-day T-bill rate in 1987 (5.775 percent),
giving $13,283 million. This figure excludes coin and currency held by the U.S.
Treasury and Reserve Banks but includes
idle cash balances at depository institutions. That vault cash at banks can be used
to satisfy reserve requirements reflects the
fact that effective reserve requirements
were lowered in 1959, not the possibility

FEDERAL RESERVE BANK

63

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

and government payor costs. According to
U.S. Postal Service sources, there were
153.9 billion pieces of mail handled in fiscal 1987. Earlier analyses by the University of Michigan’s Survey Research Center
(Household Mail Stream Study, Prepared for
the Mail Classification Research Division,
Rates and Classification Department, U.S.
Postal Service, 1978; and Nonhousehold
Mail Stream Study, Prepared for the Mail
Classification Research Division, Rates and
Classification Department, U.S. Postal Service, 1980) indicated that 82.5 percent of
all mail originated in the nonhousehold
sector, whereas 17.5 percent originated
from households. Of nonhousehold-originated mail, 35 percent was bill related and
typically included checks sent to pay for
bills received. In contrast, 75 percent of all
household-originated mail was bill related.
Overall, some 42 percent (0.75 × 0.175 +
0.35 × 0.825 = 0.42) of all mail (153.9 billion items) is estimated to be payment related—bills sent for collection and checks
sent for payment. At $0.24 each, the
64,638 million in payment-related items
generates $15,513 million in consumer,
small business, and corporate payor stamp
and envelope costs. Because almost all of
these mailed items are likely to be first
class mail and the number of first class
items was 78.9 million in 1987, bill payment–related mail comprises 82 percent of
all first class mail in this estimate.
Bank processing and transportation
costs per check were estimated at $0.32,
which includes (1) costs of crediting a deposit account of $0.057 per credit, (2)
costs of processing and transporting transit items (either by the payee bank or its
intermediaries) of $0.049 per item, (3)
costs of an on-us debit of $0.177 per debt,
(4) costs of handling return items of
$0.012 per item, and (5) the cost of returning canceled checks to account holders of $0.029 per check. These estimates
were based primarily on data from the
1986 Functional Cost Analysis data for the
76 banks with more than $200 million in
deposits. (See Board of Governors of the
Federal Reserve System. Functional Cost
Analysis, 1986.) The smaller banks were

discarded to offset the bias toward smaller
banks in the FCA sample, although the estimates would not be substantially different if these smaller banks were included.
The estimate in (1) was determined by dividing the cost of handling a deposit
($0.3627) by the number of checks per average deposit (6.386). The estimate in (2)
was determined by multiplying the cost of
a transit check deposited ($0.0975) by the
proportion of total handlings accounted
for by transit items (0.507). (Note that
this proportion is less than the proportion
of transit checks overall, because all transit
checks are also handled as on-us items by
other banks.) The estimate in (3) is simply
the FCA’s estimate of the cost of an on-us
debit. The estimate in (4) was determined
by using data from a return item study by
J. D. Carreker and Associates. (See Carreker, J. D., and Associates, Inc. Return
Item Study: Final Report. Prepared for the
Bank Administration Institute, 1985.) It
was estimated that 350 million out of
about 40 billion items in 1985 were returned—a ratio of 0.00875. The estimated
costs to reject the item are $0.71, and the
estimated cost to send the item each step
backward through the endorsement chain
is $0.43. We assume that the $0.43 also
applies to returning the item to the payee.
The average return-item cost of $0.012
was therefore determined to be (0.00875)
× {$0.71 + $0.43 + [$0.43 × 0.507]},
where 0.507 is the transit-handling proportion discussed previously. The estimate
in (5) was determined by taking the check
safekeeping savings of $7 per account per
year (that is, the savings from not returning canceled checks to payors) reported by
Valley National Bank. (See Wall Street Journal. “Canceled Checks Are Destroyed at
Some Banks, Producing Savings.” February
21, 1985 p. 1.) and dividing by the number of checks written per account per year
(237.28). Thus bank processing costs are
$0.32 × 47 billion checks = $15,040 million. Together, the business and government payor accounting and disbursing
costs ($5,088 million), the mailing costs
for all payors ($15,513 million), and the
bank costs just derived total $35,641 mil-

FEDERAL RESERVE BANK

64

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

lion in check-processing costs. Bank costs
will include Federal Reserve check-processing and transportation costs.
d

2-A1, pp. 77–9, divided by 365 days, multiplied by the 45 days a transaction is outstanding, multiplied by the 90-day Treasury
bill rate of 0.05775.]

Check

ACH

g

The total value of check float in 1987
is estimated at $39.1 billion. [See Humphrey and Berger (1990) Table 2-2, pp. 54–5,
note c]. Dividing by 47 billion checks
written gives a per-item float benefit of
$0.83. The cost of holding funds earning
little or no interest in a checking account
is assumed to be a soft-dollar payment for
services and is therefore implicitly included under production and processing
costs rather than float costs.
e

Payor ACH costs are estimated to be
$0.18 per ACH item, based on the government’s direct deposit ACH costs. (See Dudley, William C. A Comparison of Direct
Deposit and Check Payment Costs, ed. 2.
Board of Governors of the Federal Reserve
System, 1983.) Applying this origination
cost to the total volume of ACH items
originated [936 million items in Humphrey and Berger (1990) Table 2-A1, pp.
77–9] gives $168 million as an estimate
of government and business payor costs.
Payee and bank ACH costs were derived
by multiplying the 1987 weighted average
per item commercial bank price for ACH
($0.089, from Trans Data Corporation.
1988 Bank Pricing Program, 1988, p. 42)
by 1987 ACH item volume (936 million),
giving $83 million. This includes per-item
fees, tape-handling costs, and other ACH
charges. The commercial bank prices used
included all Federal Reserve costs [$71
million, from Board of Governors of the
Federal Reserve System. Planning and Control System Expense Report (PACS), 1987,
including a 16 percent PSAF]. Last, some
bills paid through an ACH are first mailed
to payors as a notification before debiting a
customer’s account. These costs ($0.22 for
postage plus $0.02 for an envelope) are assumed to apply to 20 percent of ACH commercial volume, which is around one-half
of total ACH volume of 936 million. Thus
mail costs of $22 million ($22 million =
$0.24 × 0.10 × 936 million) are added to
ACH costs of $251 million, giving total
ACH processing costs of $273 million.

Credit Cards

Credit card production costs—which
include the costs of issuing the cards,
maintaining accounts, and paying merchants—on average equaled the cost of
float. (See Bank Administration Institute.
Banking Issues and Innovations, 1985. This
was referenced in American Banker, April
9, 1985, p. 16.) Production costs are presumed equal to the cost of float today as
well ($2,257 million). Retail or merchant
processing costs were $0.44 per credit card
transaction, giving a total processing cost
of $2,249 million [from 5,111 million card
transactions in Humphrey and Berger
(1990) Table 2-A1, pp. 77–9, multiplied
by $0.44, from Board of Governors of the
Federal Reserve System. Credit Cards in the
U.S. Economy, 1983, p. 43]. Thus the total
social (user) costs of credit card transactions are estimated at $4,506 million
($2,249 million) with a unit cost estimate
of $0.88 ($0.44) per transaction.

Credit Cards

f

According to a Bank Administration Institute study (Banking Issues and Innovations,
1985), bank credit card transactions are outstanding an average of 45 days. Thus total
credit card float costs are estimated at
$2,257 million. [This estimate is based on a
figure of $317 billion in yearly charge volume in Humphrey and Berger (1990) Table

h

ACH

ACH debits do create some float, like
checks, but there is no float associated with
ACH credits, which are like European giro
payments. ACH debit float, evaluated at the
90-day Treasury bill rate cited previously,
gives less than $1 million in float value in

FEDERAL RESERVE BANK

65

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Department, 1978, p. 32. Penzer estimated
that a traveler’s check was outstanding for
an average of 57 days in 1976.) Float revenues to the issuer are $525 million [from
0.05775—the 90-day Treasury bill rate in
1987—multiplied by $7.0 billion and divided by (l - 0.23)], which was the average
daily value of outstanding nonbank traveler’s checks. [See Humphrey and Berger
(1990) Table 2-A1, pp. 77–9, note d, adjusted upwards to account for bank traveler’s checks.] Thus the costs of issuing and
paying traveler’s checks, which would also
cover the expense of funding lost checks,
is $525 million plus $470 million for a
total of $995 million. The retail cost of
handling and processing traveler’s checks
is assumed to be equal to that for cash of
$0.45 per transaction (Board of Governors
of the Federal Reserve System, 1983,
Credit Cards in the U.S. Economy, Washington, D.C., p. 43), which yields a cost of
$0.45 times 1,345 million transactions
(Humphrey and Berger, 1990, Table 2-A1,
pp. 77–9) = $609 million.

1987. Wire transfer also can create some
float and fluctuates between debit and
credit float [none for Clearinghouse Interbank Payments System (CHIPS)]. Over
1987, wire transfer debit float valued at
less than $2 million was created. Debit
float is a user benefit; credit float is a user
cost.

POS (Debit cards)

i

Some POS networks use a direct debit
to an account (like an ATM withdrawal or
bill payment with a unit cost of $0.66).
Other networks are configured to work
more slowly through an ACH, which has
a unit cost of $0.27. Lacking strong evidence on the real cost of POS, we assumed
that it is likely to fall between that for an
ATM direct debit and an ACH transfer.
The simple average of these two unit costs
was $0.47, which was used to approximate
the cost of a POS transaction.

POS (Debit cards).

j

There is no float associated with POS
transactions (except for those functioned
through an ACH, but the value of this
float is small), so the user costs and the
social costs ($0.47 × 55 million POS
transactions = $26 million) are the
same.
k

Traveler’s Checks

l

Though there is float associated with
traveler’s checks ($525 million in float divided by 1,345 million transactions = $0.39
in float cost per traveler’s check), we have
assumed that all float revenues in fact
cover operating costs, so float in the same
sense as check float, or the opportunity
cost of holding idle funds, in the same
sense as applied to the issuance of cash by
the government, does not exist. Put differently, traveler’s check float is not the same
thing as a redistribution of income between
payor and payee but rather an alternative
method of covering operating expenses for
the issuer of the traveler’s check. Because
this is a reasonably competitive industry,
we have assumed that there is no monopoly power on the part of the issuer to obtain above-normal profits (such a situation
does not apply to the issuance of currency
by the government). In sum, the user and
social unit costs of a traveler’s check are
the same at $1.18 per transaction [($995
million plus $609 million) divided by
1,345 million transactions].

Traveler’s Checks

Production, processing, and operating
costs for traveler’s checks are proprietary.
Thus it was assumed that revenues associated with issuing traveler’s checks equaled
the costs involved. Generally, an issuing
fee of 1 percent of the face value of the
traveler’s check is assessed, yielding a revenue flow of $470 million [from 0.01 multiplied by $47 billion, in Humphrey and
Berger (1990) Table 2-A1, pp. 77–9]. More
revenue for the issuer is obtained from
float because it is estimated that the average traveler’s check is outstanding for 70
days. (See Penzer, Michael L. “The Nature
and Size of Money Order and Traveler’s
Check Markets in California and the Nation.” Economic and Financial Information Division, California State Banking

FEDERAL RESERVE BANK

66

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Postal Service, Costs and Revenue Analysis
Report, 1987). Revenues, however, were
$1.15 per item ($148 million from fees
plus $16 million from float, giving $164
million total, divided by 142 million
items). Revenues from money orders
($164 million) in excess of directly allocated costs ($112 million) are allocated to
Postal Service overhead for all services offered. These overhead costs are viewed as
joint costs and are reallocated back to the
individual services according to certain
criteria, one of which is the value of the
service to the user (Ramsey pricing). Thus
it is impossible to determine whether
postal money order revenues cover all
costs or if excess profits (or losses) are
being incurred and cross-subsidization exists. Penzer has noted that use of postal
money orders peaked in 1952 and subsequently lost market share to bank-issued
money orders and private firms. (See Penzer, Michael L. “The Nature and Size of
Money Order and Traveler’s Check Markets in California and the Nation.” Economic and Financial Information Division,
California State Banking Department,
1978, p. 6.) This implies that postal
money orders face a competitive market
and, if anything, may be cross-subsidized
rather than be used to cross-subsidize
other postal services. As a result, we assume here that all postal money order revenues are used to cover all costs, even
though it is likely that costs may exceed
total revenues by some unknown amount.
This implies that the fully allocated cost of
a postal money order is at least $1.15 per
item, and this figure is used to approximate the unknown production cost of all
money orders. Thus total production costs
of all money orders is estimated at $1.15 ×
(811 million items) = $933 million, for a
total social cost of $1,451 million. Total
user costs are the same because total
money order float costs of $89 million
(from $0.11 multiplied by 811 million
items) are presumed fully used to cover
real resource costs and represent an alternative charging method rather than a
transfer payment.

Money Orders

m

Federal Reserve cost in processing 146
million postal money orders in 1987 was
$2.8 million [direct and allocated costs
plus overhead (39 percent of total activity
costs), from Board of Governors of the
Federal Reserve System. Planning and Control System Expense Report (PACS), 1987],
or $0.019 per money order. This unit cost
figure was applied to the 811 million
money orders estimated to have been used
in 1987 [Humphrey and Berger (1990)
Table 2-A1, pp. 77–9], giving $15 million.
Merchant or receiver processing costs were
assumed to equal those reported for
checks at a sample of retail stores: of
$0.50 per item for a total merchant processing cost of (811 million money orders)
× ($0.50) = $406 million. (See Board of
Governors of the Federal Reserve System.
Credit Cards in the U.S. Economy, 1983, p.
43.) It was assumed that one-half of all
money orders are mailed by the user, giving an extra user cost of $97 million [from
per-item postage cost ($0.22) plus per-envelope cost ($0.02) multiplied by 0.5 and
then multiplied by 811 million items =
$97 million]. This assumption is supported in a survey of money order users in
California by Pierce, who found that the
payment of utility and other monthly bills
plus sending money to relatives or friends
accounts for almost two-thirds of the responses as to why money orders were
used. (See Pierce, James L. “The Users of
Money Orders,” University of California–
Berkeley, 1977. Appendix Table A-3, p. 8.)
Total money order processing costs are
thus estimated at $518 million. Money
order production costs, including all costs
of distributing them to users plus the costs
of redemption, are taken from postal
money order fiscal year data for 1984 (and
thus will not match exactly with the volume and value figures for postal money
orders processed by the Federal Reserve in
calendar year 1987). The directly allocated
production costs for postal money orders
were $0.79 per item ($112 million in
directly allocated cost divided by 142
million items in fiscal year 1987 from U.S.

FEDERAL RESERVE BANK

67

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

n

Money Orders

Money orders are estimated to be outstanding between 5 and 11 days. (See Penzer, Michael L. “The Nature and Size of
Money Order and Traveler’s Check Markets in California and the Nation.” Economic and Financial Information Division,
California State Banking Department,
1978, p. 8.) Taking the midpoint (8 days)
generates an estimated float benefit of $16
million for postal money orders ($16 million = $12.511 billion annual value of
postal money orders outstanding multiplied by 8/365 of a year multiplied by an
interest rate of 0.05775).

Wire Transfers

o

Wire transfer volume in Humphrey
and Berger (1990) Table 2-A, pp. 77–9 (84
million) was multiplied by a weighted average commercial bank charge for wire
transfers ($7.33, from Trans Data Corporation. 1988 Bank Pricing Program, 1988,
p. 72) to give the total cost of wire transfers ($616 million). The Federal Reserve
wire transfer cost components [from Board
of Governors of the Federal Reserve System. Planning and Control System Expense
Report (PACS), 1987] and the PSAF were
$66 million, which was assumed to be
fully passed to final users in these bank
fees.
p

Wire Transfers

Wire transfer also can create some
float, which fluctuates between debit and
credit float (none for CHIPS). Over 1987,
wire transfer debit float valued at less than
$2 million was created. Debit float is a
user benefit; credit float is a user cost.

FEDERAL RESERVE BANK

68

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Yang Woo Kim is a senior economist for the Bank of Korea. From August 1995 to July 1996, he was a visiting scholar in the research
department of the Federal Reserve Bank of St. Louis. Nick Meggos and Kelly Morris provided research assistance. The views expressed are
those of the author and do not necessarily reflect official positions of the FRB of St. Louis, the Federal Reserve System, the Board of
Governors, and the Bank of Korea. The author is grateful to Mike Dueker, Bill Gavin, Barry Jones, and Travis Nesmith for helpful comments
and suggestions.

post-Depression period, however, with
the possible exception of a period in the
1950s or 1960s. These facts suggest that
procyclical price movement is not a
stable feature across many business
cycles.
Chadha and Prasad (1993, 1994) show
the importance of making a clear distinction between inflation and the cyclical
component of the price level in reports
and interpretations of stylized facts regarding business cycles. They find that, in
postwar quarterly data for the United
States, as well as for the rest of the G-7
countries, the cyclical components of the
price-level and output series are negatively
correlated. The inflation rate, however, is
generally correlated positively with various
measures of the cyclical component of
output, suggesting that demand-driven
models of the business cycle are not necessarily falsified by the countercyclical
behavior of the price level.
Similarly, Hall (1995) argues that the
conclusions on the sources of the business
cycle of Cooley and Ohanian, and Kydland
and Prescott are flawed. He shows that a
negative correlation between detrended
output and detrended prices is predicted
by the natural-rate nominal-demand shock
model, under reasonable assumptions. In
other words, he shows that the naturalrate model suggests that procyclical prices
imply a positive correlation between the
change in inflation and detrended output.
Judd and Trehan (1995) argue that the use
of cross-correlation coefficients as indicators for evaluating the empirical relevance
of demand-oriented versus supply-oriented
macroeconomic theories is problematic in
principle. Using two econometric models,
they show that correlation coefficients
for prices and output could easily be negative even if demand shocks were the
primary source of cyclical fluctuations
and prices were procyclical. Gavin and
Kydland (1995) show that alternative
money supply rules can change the

Are Prices
Countercyclical?
Evidence From
East Asian
Countries
Yang Woo Kim
The cyclicality of prices has been in
debate recently. One of the key stylized
facts that has served as a cornerstone for
traditional business cycle models is the
procyclical behavior of prices. Recent
research has questioned the empirical basis
for this stylized fact. Using quarterly U.S.
data, Kydland and Prescott (1990) report
that cyclical components of U.S. prices
and output are correlated negatively in
the post–Korean War period. They cite
this result as evidence against the conventional understanding that prices are procyclical.
Cooley and Ohanian (1991) confirm
the negative correlation in the post-WWII
period and also find that if the sample is
extended into the nineteenth century, the
data are not at all suggestive of procyclical
prices, with the important exception of the
inter-war period. They interpret their findings of countercyclical prices as being
inconsistent with the predictions of demanddriven models and instead supportive of
real business-cycle models.
The work of other researchers, such
as Backus and Kehoe (1992) and Smith
(1992), complements these studies by
examining cross-country evidence. For
example, Smith finds that, for ten countries, fluctuations in the price level
generally are procyclical from the late
nineteenth century until WWII, with the
exception of a period around WWI. Pricelevel movements are countercyclical in the

FEDERAL RESERVE BANK

69

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

cyclical nature of prices in a flexible-price
economy.
The purpose of this paper is to document the empirical evidence on the
cyclical behavior of prices and inflation
in two rapidly growing developing countries—Korea and Taiwan. I use the methodology of Chadha and Prasad (1994) to
see if results from advanced countries are
robust when compared to the results from
these developing countries, whose business cycles may have different characteristics. In addition to the usual techniques
for detrending—linear detrending, first
differencing, and Hodrick-Prescott (HP)
filtering—two more techniques are utilized.
That is, the Zivot-Andrews’ (1992) procedure of a unit root test with a structural
break at an unknown point in time is used
to enable researchers to consider possibilities of a segmented trend of the data. Also,
a smooth trend based on the phase average
trend (PAT) technique (Klein and Moore,
1985; Zarnowitz, 1992) that is popular in
growth-cycle literature is estimated and
used in detrending the time series to confirm the robustness of the empirical
findings.
The next section of this paper discusses the strategy of the trend-cycle
decomposition and related issues. The
third and fourth sections present empirical
correlations of prices and inflation with
the cyclical component of output in Korea
and Taiwan. The final section summarizes
the main results and implications.

There are several widely used transformations, including deterministic linear
detrending, first differencing, and other
types of stochastic detrending such as
Hodrick and Prescott (1981) filtering.
Since each technique suggested has its
own limitations, depending on the nature
of the input series and the issues addressed, it is likely that different detrending methods will be useful in different
applications.
To get around problems associated
with particular detrending methods, some
researchers adopt an eclectic strategy in
detrending a time series—i.e., instead of
using a particular technique for detrending, they try various detrending methods
and evaluate the robustness of the results
across techniques. Cooley and Ohanian
(1991) and Smith (1992) use three filters
in calculating the cross-correlations of
price and output: linear, first-difference,
and HP. Chadha and Prasad (1993,1994)
try a segmented trend based on Perron
(1989) and a stochastic detrending method
used by Beveridge and Nelson (1981) and
Blanchard and Quah (1989) as alternatives.
In this study, I basically follow Chadha
and Prasad’s strategy in using a variety of
detrending procedures. I use the three filters—linear, first-difference, and HP—as
well as two others. Instead of BeveridgeNelson detrending (1981) or Blanchard
and Quah’s method (1989), I investigate
the nature of the nonstationarity of the
time series more thoroughly by applying
Zivot and Andrews’ unit root test (1992)
under the alternative of trend stationarity
with a break at an unknown point in time.
A break point identified by this procedure
is then used for setting up a segmented
linear trend. I also calculate a fifth decomposition with the Growth Cycle program
based on the PAT technique.
A deterministic linear trend is free of
any cyclical or any other short-run movements; as a result, it guarantees a clear
decomposition of the series into a trend
and cyclical component. But it is not likely
that a deterministic trend, or trendstationary process, would represent the

Trend-Cycle Decomposition

1

Precisely speaking, if a nonstationary stochastic process can
be made stationary by a particular transformation, and a different transformation is
employed, the spectral representation will be altered. Thus
it is important to use the correct transformation.

To examine the co-movement of macroeconomic variables over the business
cycle, it is necessary to use a procedure
based on an appropriate measure of the
cyclical component of the series. It has
been well documented that macroeconomic variables are nonstationary in
general. The characterization of the form
of the nonstationarity is very important,
since it will affect the nature of the
stationary component as well as the
permanent component of a time series.1
However, there is no consensus on the correct trend-cycle decomposition method.

FEDERAL RESERVE BANK

70

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

real world across periods of dramatic
changes such as the Great Crash of 1929,
or World War II or periods of structural
changes in developing economies. Another
detrending method, first differencing, is
included for completeness, but it is probably not appropriate for studying the
cyclical nature of prices, since it removes
information at cyclical frequencies. Most
real business-cycle researchers have extensively used the Hodrick and Prescott filter
(1981). Kydland and Prescott (1982)
choose it because it focuses on fluctuations
at cyclical frequencies. This technique,
however, has been criticized by several
researchers (King and Rebelo, 1993;
Cogley and Nason, 1995; Harvey and
Jaeger, 1993; Gregory and Smith, 1995),
because mechanical detrending based on
the HP filter could lead investigators to
report spurious cyclical behavior. The last
two methods I use—the segmented linear
method and the PAT technique—have not
been widely used in the business-cycle literature. Therefore, I include a detailed
discussion of these two methods before
analyzing the cyclical behavior of prices in
Korea and Taiwan.

1992). By treating the breakpoint as endogenous, they find less evidence against
the unit root hypothesis than Perron finds
for many of the data series but stronger
evidence against it for several of the series.
Similarly, Banerjee, Lumsdaine, and Stock
(1992) examine the unit root hypothesis
on postwar real output of G-7 countries
using tests based on asymptotic distribution theory, which treats break dates as
unknown a priori, and they find that the
null hypothesis of a unit root can be
rejected only for Japan. Serletis (1994),
who bases his research on the Zivot and
Andrews’ technique (1992), finds that the
unit root model can be rejected for real per
capita output series in eight out of ten
countries over the period of 1870 to 1985,
if a one-time break is allowed at an
unknown point in time. Using the same
technique, Alba and Papell (1995)
examine unit roots in aggregate and real
per capita GDP for nine newly industrializing countries (NIC) and newly exporting
(NEC) countries in east and southeast
Asia. In 15 out of 18 cases, they reject the
unit root hypothesis in favor of the theory
of trend stationarity with a break. Overall,
the results from these studies suggest that
the international evidence of the unit root
model is mixed as far as output is concerned,
either real GDP or per capita GDP; the
outcome depends on how certain big
shocks (such as the Great Crash of 1929)
are treated.
Here I use Zivot-Andrews’ variant of
the sequential Dickey-Fuller test of a unit
root, which can be represented by the following equation:

The Segmented Linear Method
Since Nelson and Plosser (1982), the
popular view has been that most macroeconomic variables have a unit root—that
is, they follow stochastic trends. Recently,
evidence against this unit root hypothesis
has been presented by various authors.
Perron (1989) shows that, once a break in
the trend is allowed in the data, the unit
root hypothesis can be rejected in favor of
the alternative hypothesis of stationarity
around a deterministic trend that has an
exogenous break at either the Great Crash
or the 1973 oil shock. This situation is
true for postwar quarterly U.S. real GNP as
well as for eight of the eleven macroeconomic time series Nelson and Plosser
examined. Perron’s results are criticized and
further elaborated by several researchers, on
the ground that his choice of a break point
in the trend is ad hoc and hence cannot
circumvent the problem of data-mining
(Christiano, 1992; Zivot and Andrews,

(1)

∆Yt = µ 1 u DUt 1 b t 1 g DTt
1 a Yt-1 1 ∑ci ∆(Yt-j) 1 et

where DU = 1 and DT = t - TB if t > TB and
0 otherwise. This tests the null hypothesis
of a unit root against the alternative
hypothesis of a trend stationarity with a
one-time break (TB) in the intercept and
slope of the trend function at an unknown
point in time. For this, different regressions are run for TB = 2, 3, ..., T-1, where
T is the number of observations adjusted

FEDERAL RESERVE BANK

71

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Table 1

Test for Unit Root: Zivot-Andrews’ Procedure 1992
Variables
Sample
Break Point
a
t-value
Model

k

Korea
GNP
Deflator
CPI

1970.1Q-95.2Q
1970.1Q-95.2Q
1970.1Q-95.2Q

1979.2Q
1979.4Q
1979.3Q

GDP
Deflator
CPI

1961.1Q-95.3Q
1961.1Q-94.4Q
1959.1Q-95.4Q

1977.4Q
1973.1Q
1982.2Q

20.25
20.22
20.15

24.6
25.41*
26.20*

A
C
C

2
5
12

24.21
24.21
23.4

B
A
B

10
4
6

Taiwan

20.23
20.06
20.05

*Significant at the 5 percent level.

2

The results of the standard augmented Dickey-Fuller unit test
suggest that output and price
series in both countries have a
unit root if no break in the
trend is taken into account.

3

The results for the unit root test
for the output series are slightly
different, in terms of either
model chosen or the break
point, compared to Alba and
Papell (1995). This difference
in outcome seems to stem from
differences in the data sets.
Alba and Papell use annual
data collected and adjusted by
Summers and Hester
(1953–1988 for Korea, 1950
–1988 for Taiwan), while I
use quarterly data pulled directly out of the IFS from IMF and
DRI. Part of the reason for the
difference will be the length of
the sample period. My data has
been updated until recently and
hence has a longer sample period than theirs, which ends at
1988.

Taiwan output series follows model B, in
which the slope of the trend changes. In
the case of prices, model C is chosen for
both the CPI and GNP deflator in Korea,
while models A and B are chosen for the
Taiwan deflator and CPI respectively. Note
that the chosen potential structural break
period in the trend of each series is not the
same for the two countries3 (see, also, Figures 1a through 1d on the time paths of
the Dickey-Fuller t-statistics). This difference suggests that a break-inducing
exogenous event, such as the oil shock in
the 1970s, may not have affected these
countries at the same time, partly because
of differences in government policy and
different levels of development in each
country. Among the six cases, the unit root
hypothesis is rejected for the Korean deflator and CPI at the 5 percent level of
significance. The results from Table 1 may
imply that the nature of the nonstationarity of the output series is different
country-by-country; therefore, care should
be taken in applying a specific detrending
procedure.

for lost data caused by differencing and lag
length k. The lag length is selected according to the procedure suggested by
Perron (1989). Working backwards from
k = 12, I choose the first value of k such
that the t-statistic on the last lag in the
autoregression is greater than 1.6 in absolute value and the t-statistic on the last
lag in higher order autoregression is less
than 1.6.
A break point is chosen that gives the
least favorable result for the null hypothesis and the most weight to the trendstationary alternative. This result can be
accomplished by choosing the minimum
t-statistic on the Dickey-Fuller statistic out
of T-2 regressions. Also, following Perron
(1989), I tried three possible models under
the alternative hypothesis to test the unit
root: model A, which allows a break in the
intercept (DUt); model B, which allows a
break in the slope only (DTt); and finally
model C, which allows a break in both
slope and intercept. Model C, which is represented by Equation 1, is used if the
t-statistics on both the intercept and the
slope dummies are significant. If not,
either model A or B is used, depending
upon which dummy from the model C is
significant.
Table 1 shows the results for a unit
root hypothesis of output and prices for
Korea and Taiwan based on quarterly
data.2 The Korean output series conforms
with model A, implying that there is a
change in intercept of the trend, while the

The Phase Average Trend Technique
Another trend-fitting measure used is
the Phase Average Trend (PAT) technique,
combined with the turning-point selection
program, which is used in growth-cycle literature. Growth cycles are measured by
calculating the observed monthly deviations from the trend. Klein and Moore
(1985) have adapted the computer pro-

FEDERAL RESERVE BANK

72

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Figure 1

Dickey-Fuller t-statistics
1a: Korea GNP

1b: Korea CPI

0

0.8
0.0

21

20.8
21.6

22

22.4
23.2

23

24.0
24.8

24

25.6
25

70

73

76

79

82

85

88

91

94

1c: Taiwan GDP

26.4

73

76

79

82

85

88

91

94

1d: Taiwan CPI

21.0

21.50

21.5

21.75
22.00

22.0

22.25

22.5

22.50

23.0

22.75

23.5

23.00

24.0
24.5

70

23.25
59

63

67

71

75

79

83

87

91

gram for dating classical cycles, originally
developed by Bry and Boschan (1971), so
that it can be used in producing growthcycle chronologies. The traditional way of
getting a trend from the data in early business-cycle literature is to calculate the
moving average of the series for a fixed
period of time in the context of the beginning and end points of the data. Klein and
Moore find the trend rate calculated in this
way to be noticeably affected by the shorter cycles in the series. To correct for this
problem, they devised the new trendfitting technique, PAT.
The PAT technique involves a trend
adjustment—a moving average over the
two or three phases of cycles of varying

95

23.50

59

63

67

71

75

79

83

duration. For example, after the seasonally
adjusted data are smoothed to produce a
seventy-five–month moving average, observed deviations from this initial trend are
first calculated. From this rough deviation
cycle, tentative turning points (peaks and
troughs) are identified and temporary
phases of varying lengths are determined.
Then final trend estimates are computed as
a two- or three-phase moving average of
the original, seasonally adjusted data,
interpolating monthly between the centered values of these averages. The cyclical
component of the series can be obtained
by calculating the deviations of the original data from this refined and flexible
nonlinear trend estimate. Klein and Moore

FEDERAL RESERVE BANK

73

OF

S T. L O U I S

87

91

95

S E P T E M B E R /O C T O B E R 1996

Figure 2

Detrended Output: Korea
a.
0.15

HP-Filtered

0.10

PAT-Detrended

Seg. Detrended

0.05
0.00
20.05
20.10
20.15

70

72

74

76

78

80

82

84

86

88

90

92

94

b.
0.15

Linear Detrended
Differenced

0.10
0.05
0.00
20.05
20.10
20.15

70

72

74

76

78

80

82

argue that the trend-fitting method will be
more satisfactory than the simple moving
average in separating cyclical influences
from the underlying trend. As a result,
they argue, the data can then identify
relatively stable trend rates of growth,
unassociated with the shorter cycles—
a primary concern of business-cycle researchers. This procedure is applied to the
Korean and Taiwan data to obtain the
alternative measure of cyclical components.

84

86

88

90

92

94

difference method suggests too many to be
consistent with conventional ideas about
the length of business cycles.
To identify the similarities of each calculated cyclical component, correlations
between measures are calculated in Table
2. The lower triangle of each panel represents correlations between cyclical components of output, while the higher triangle of each panel represents correlations
between cyclical components of price calculations based on the consumer price
index. Note that the correlations between
the first-differenced series and those detrended by either linear trend or segmented
trend are not significant. This is the
expected result from Table 1, because
output and prices should be modeled as
having either a deterministic or a stochastic trend.
In general, the cyclical components of
output and prices, when detrended by the
PAT technique, are closely related with the
cyclical components derived from all other
methods. In particular, PAT detrended
components are most strongly related with

Comparing the Alternative
Measures of the Cyclical Component
Figures 2a and 2b depict the five alternative measures of the cyclical component
for Korean GNP. The three most similar,
shown in Figure 2a, are those derived
through the HP, the PAT, and the segmented linear trend methods. For all three
methods, the average frequency and amplitude of the calculated cycles appears to be
roughly consistent with business cycles
lasting 4 to 6 years. Figure 2b includes the
two extremes: The linear trend method
suggests too few cycles, and the first-

FEDERAL RESERVE BANK

74

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Table 2

Correlations between Cyclical Components of Output and Prices
KOREA
Output Prices

Linear Trend

Segmented Trend

HP Filter

PAT

Linear trend
Segmented trend
HP filter
PAT
First Difference

1
0.69
0.76
0.76
0.23

0.3
1
0.74
0.75
0.18
TAIWAN

0.37
0.69
1
0.97
0.39

0.66
0.52
0.81
1
0.4

Linear trend
Segmented trend
HP filter
PAT
First Difference

1
0.53
0.49
0.6
0.15

0.8
1
0.91
0.83
0.24

0.45
0.52
1
0.96
0.31

0.67
0.8
0.9
1
0.28

Difference
0.08

20.13
0.17
0.2
1
0.1

20.01
0.22
0.17
1

* The lower (higher) triangle of each panel represents the correlation between cyclical components of output (prices).

HP filtered series. For example, the correlations between the two measures in Korea
are, respectively, 0.97 for output and 0.81
for prices. The cyclical components calculated with the first-difference filter are least
correlated with the cyclical components
calculated by means of the other methods.

implications, because first-differenced
series do not generally yield the cyclical
components of the original series.
Table 3 shows the cross-correlations
of prices and output, to each of which the
same transformation techniques are applied. For each country, two measures of
price series are used. The correlations are
reported up to four lags and leads, and the
standard errors are computed under the
null hypothesis that the true correlation
coefficient is zero. The results show that
the standard errors for cross correlation
coefficients are very similar.
The first panel reports the correlations
between prices and output series, with
both series detrended using a simple linear
trend. The numbers are all significantly
negative for Korea. However, in the case of
Taiwan the correlations are positive at all
leads and lags with the deflator, and
mostly positive with the CPI.
The second panel shows the result
when both prices and output series are
modeled as stationary around a segmented
time trend, allowing for a different break
in each series. The correlations in Korea
are still negative, although smaller than
those of the first panel. In Taiwan, however, the correlations change their signs to
become negative for both the deflator and
the CPI.

Correlation Between Cyclical
Components of Prices and Output
In this section, price-output relationships for each country are investigated by
means of various detrended series based
on the techniques discussed above. First,
I detrend prices and output using a deterministic linear trend. Second, I detrend
prices and output using a segmented linear
trend, with a break point in either the level
and/or slope of the trend, depending on
which model is appropriate for each
country. The periods for a structural break
in the trend of output and prices are different across the countries. Third, I use the
HP filter to detrend the prices and output
series. Fourth, I detrend prices and output
by means of the PAT technique. And
finally I take first differences of both prices
and output. By construction, these are
interpreted as the relation between
inflation and economic growth rates. As
discussed above, care should be taken in
interpreting these results as having cyclical

FEDERAL RESERVE BANK

75

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Table 3

Cross Correlations of Prices and Output with Same Detrending
Methods for Both Series
Lead/Lag
(2 denotes lag)

Korea
Deflator

Taiwan
CPI

Deflator

CPI

Detrended Prices and Output (Linear Trend for Both)
4
3
2
1
0
21
22
23
24

20.50
20.50
20.49
20.47
20.45
20.41
20.38
20.34
20.31

(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)

20.59
20.61
20.61
20.60
20.58
20.55
20.51
20.47
20.42

(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)

0.12
0.15
0.19
0.23
0.29
0.33
0.38
0.42
0.45

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)

20.02 (0.09)
20.00 (0.09)
0.02
0.06
0.11
0.16
0.21
0.27
0.32

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)

20.53
20.55
20.56
20.54
20.49
20.41
20.32
20.24
20.15

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)

Detrended Prices and Output (Segmented Trend for Both)
4
3
2
1
0
21
22
23
24

20.15
20.09
20.18
20.19
20.29
20.31
20.32
20.22
20.18

(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)

20.18
20.26
20.34
20.42
20.50
20.53
20.43
20.35
20.21

20.14
20.20
20.26
20.29
20.29
20.29
20.26
20.23
20.23

(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)

Detrended Prices and Output (HP Filter for Both)
4
3
2
1
0
21
22
23
24

20.29
20.32
20.39
20.40
20.47
20.37
20.26
20.15
20.07

(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)

20.26
20.39
20.51
20.57
20.58
20.54
20.48
20.37
20.21

20.68
20.76
20.74
20.63
20.47
20.27
20.08

(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)

The third panel shows the results for
the HP-filtered prices and output. The
basic picture remains the same as in the
second panel, but negative associations are
getting stronger in Korea (see also Figure
3, page 78) and at the leads in Taiwan. The
fourth panel shows that the correlations
are generally negative for both countries
when the cyclical components of price and
output series are extracted by PAT tech-

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
0.07 (0.09)
0.20 (0.09)

20.59
20.73
20.79
20.75
20.63
20.45
20.26
20.09

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
0.09 (0.09)

nique. A striking feature of this panel is
that in Taiwan negative correlations stand
out and are stronger than in Korea, especially at the leads.
Finally, the correlations between the
first log-differenced prices and output (i.e.
inflation and economic growth) are shown
in the last panel. Again they are all negative in Korea except lead 1 and lag 3 in
the deflator case, although they are weaker

FEDERAL RESERVE BANK

76

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Table 3 (con’t)

Cross Correlations of Prices and Output with Same Detrending
Methods for Both Series
Lead/Lag
(2 denotes lag)

Korea
Deflator

Taiwan
CPI

Deflator

CPI

Detrended Prices and Output (PAT for Both)
4
3
2
1
0
21
22
23
24

20.14
20.13
20.18
20.18
20.23
20.17
20.10
20.03

(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
0.01 (0.10)

20.17
20.25
20.33
20.36
20.38
20.35
20.31
20.24
20.14

(0.10)
20.55
(0.10)
20.61
(0.10)
20.58
(0.10)
20.49
(0.10)
20.35
(0.10)
20.17
(0.10)
20.00
(0.10)
0.14
(0.10)
0.26
First Differences of Prices and Output

4
20.08 (0.10)
3
20.07 (0.10)
2
20.14 (0.10)
1
0.01 (0.10)
0
20.28 (0.10)
21
20.05 (0.10)
22
20.03 (0.10)
23
0.03 (0.10)
24
20.05 (0.10)
* Figures in parentheses indicate standard errors.

20.05
20.16
20.21
20.20
20.20
20.10
20.20
20.15
20.01

20.20
20.34
20.31
20.20
20.13

(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)

0.01
0.13
0.09
0.15

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)

20.51
20.62
20.68
20.65
20.55
20.39
20.23
20.07

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)

20.21
20.28
20.32
20.29
20.23
20.04

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
0.08 (0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
0.00 (0.09)
0.01 (0.09)
0.15 (0.09)

Correlation Between Inflation and
the Cyclical Component of Output

than those by other detrending methods,
implying the negative association between
inflation and economic growth. In Taiwan,
however, correlations are negative for all
the leads and insignificantly positive at
most lags.
In sum, the implication from Table 3 is
that, whichever method is used in detrending, the cross correlations of the cyclical
components of prices and output are negative for all the cases for Korea and for most
cases for Taiwan, although the significance
of the relationship is different.4 The results
do not change much regardless of which
measure of prices, either the deflator or
CPI, is used. This indicates that there is
strong evidence for the countercyclical
price behavior in Korea and Taiwan as
in the case for the industrialized countries documented in Chadha and Prasad
(1994).

As discussed above, many conventional models of business cycles imply that
the inflation rate, rather than the price
level, is procyclical. Hence the correlations
are calculated again between inflation and
detrended output. The comparison of
inflation versus the detrended price level is
illustrated for the HP method in Figure 3.
Figure 3a shows the clearly countercyclical
nature of price and output reported in
Table 3. Figure 3b shows the procyclical
nature of inflation that is documented
below.
Figure 3b plots the inflation rate with
the deviations of output from the HP
trend. Table 4 presents correlations
between inflation and various measures of
the cyclical component of output similarly

FEDERAL RESERVE BANK

77

OF

S T. L O U I S

4

In Taiwan, detrending using a
simple linear trend may not be
appropriate, since all the other
detrending methods give similar
results.

S E P T E M B E R /O C T O B E R 1996

Figure 3

a. HP-Filtered Prices and Output: Korea
0.15

Prices
Output

0.10
0.05
0.00
20.05
20.10
20.15

70

72

74

76

78

80

82

84

86

88

90

92

94

b. Inflation and HP-Filtered Output: Korea
0.15

Inflation
Output

0.10
0.05
0.00
20.05
20.10
20.15

70

72

74

76

78

80

82

defined as in Table 3. The rate of inflation
is measured as the first difference of the
log of the price level. The remarkable features of Table 4 are that (1) the positive
correlations between inflation and output
are prevalent across all the detrending
methods, contrary to the correlations
between price level and output, which are
negative in most cases, and (2) that the
magnitudes of the correlations are larger
in Taiwan than in Korea. The first panel
shows that the correlations between inflation and output that have been detrended
using a linear trend are all positive for
Taiwan, regardless of price measures, and
for Korea they are all positive when inflation is measured by the deflator. The
second through fourth panels show that
the correlations between inflation and
output are generally positive for both
countries, no matter which method is used
to take this trend out of the data.
Finally, the correlations between inflation and cyclical output that have been
detrended by the PAT technique are again
basically positive for both countries, a

84

86

88

90

92

94

result that is very similar to that shown in
panels 2 and 3. The results in Table 4 indicate procyclical variations in inflation for
most cases, in contrast to the case when
price level is used. This finding is consistent with the findings for G-7 countries
presented by Chadha and Prasad (1994).
Table 5 presents the results for the test
of the possibilities of a change in correlation structure between prices or inflation
and cyclical output before and after the
identified break point by Zivot and
Andrews’ procedure (1992). Although
output series in both Korea and Taiwan
turn out to be difference-stationary statistically, both countries may have experienced
a structural change. Hence the same
analysis is done during the subsample
period, before and after the break point.
The break point used for each country
is that for output from the Zivot and
Andrews test (Table 1). Instead of reporting all the cross correlations, the correlation between cyclical prices and inflation
and output, based on the HP filter and the
growth cycle program, are presented.

FEDERAL RESERVE BANK

78

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Table 4

Cross Correlations of Inflation and Cyclical Components
of Output
Lead/Lag
(2 denotes lag)

Korea
Deflator

Taiwan
CPI

Deflator

CPI

Inflation and Detrended Output (Linear Trend)
4
3
2
1
0
21
22
23
24

0.06
0.10
0.13
0.19
0.19
0.31
0.34
0.35
0.34

20.11 (0.10)
20.09 (0.10)
20.02 (0.10)

(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)

0.08
0.16
0.26
0.30
0.40
0.47

0.16
0.22
0.31
0.40
0.46
0.49
0.47
0.43
0.40

(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)

0.07
0.12
0.20
0.27
0.36
0.42
0.43
0.42
0.42

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)

Inflation and Detrended Output (Segmented Trend)
4
3
2
1
0
21
22
23
24

20.00 (0.10)
20.03 (0.10)
0.03
0.14
0.10
0.20
0.23
0.20
0.14

20.09 (0.10)
20.08 (0.10)
20.00 (0.10)

(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)

0.12
0.18
0.31
0.33
0.33
0.36

20.20 (0.09)
20.09 (0.09)
0.08
0.25
0.36
0.43
0.44
0.39
0.35

(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)

20.30 (0.09)
20.19 (0.09)
20.05 (0.09)
0.11
0.26
0.38
0.41
0.41
0.41

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)

Inflation and Detrended Output (HP Filter)
4
3
2
1
0
21
22
23
24

20.08 (0.10)
20.06 (0.10)
20.04 (0.10)

4
3
2
1
0
21
22
23
24

20.04 (0.10)
20.01 (0.10)

20.25
20.27
20.20
20.08

0.04 (0.10)

20.01 (0.10)
0.18
0.21
0.21
0.17

0.03
0.14
0.18
0.31
0.41

(0.10)
(0.10)
(0.10)
(0.10)

20.26 (0.09)
20.15 (0.09)

(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)

0.06
0.25
0.37
0.45
0.45
0.37
0.32

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)

20.37 (0.09)
20.26 (0.09)
20.09 (0.09)
0.09
0.27
0.40
0.42
0.41
0.41

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)

Inflation and Detrended Output (PAT)

0.01
0.10
0.06
0.24
0.27
0.26
0.23

20.25
20.25
20.17
20.05

(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)

0.06
0.17
0.20
0.32
0.42

20.16 (0.09)
20.05 (0.09)

(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)
(0.10)

0.14
0.33
0.45
0.53
0.52
0.45
0.40

Note: Standard errors are provided in parentheses.

FEDERAL RESERVE BANK

79

OF

S T. L O U I S

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)

20.29 (0.09)
20.18 (0.09)
20.02 (0.09)
0.16
0.32
0.46
0.47
0.46
0.46

(0.09)
(0.09)
(0.09)
(0.09)
(0.09)
(0.09)

5

The test for statistical significance is based on the Wald test
described in Ostle (1963),
pages 225-227. The Wald statistic is distributed as a Chisquare with one degree of
freedom. The value for a ten
percent critical region is 2.71.

S E P T E M B E R /O C T O B E R 1996

Table 5

Cross Correlations of Prices and Output: Pre- and Post- Break
Lead/Lag
(2 means lag)

Korea
Deflator

Full
PreSample Break

Taiwan
CPI

PostFull
PreBreak Sample Break

Deflator
PostFull
PreBreak Sample Break

CPI
PostFull
PreBreak Sample Break

PostBreak

Detrended Prices and Output (HP Filter for Both)
4
3
2
1
0
21
22
23
24

20.29
20.32
20.39
20.40
20.47
20.37
20.26
20.15
20.07

20.18
20.12
20.21
20.25
20.37
20.25
20.11

20.36
20.44
20.50
20.49
20.55
20.45
20.38
0.04 20.30
0.05 20.22

4
3
2
1
0
21
22
23
24

20.14
20.13
20.18
20.18
20.23
20.17
20.10
20.03

20.10
20.01
20.09
20.13
20.25
20.15
20.03

4
3
2
1
0
21
22
23
24

20.08 20.15 20.33
20.06 20.01 20.42
20.04 20.08 20.32
0.04 20.07 20.14
20.01 20.12 20.16

20.26
20.32
20.35
20.33
20.33
20.28
20.23
0.09 20.19
0.10 20.17

20.26
20.39
20.51
20.57
20.58
20.54
20.48
20.37
20.21

20.23
20.33
20.47
20.56
20.55
20.43
20.30
20.17
20.05

20.17
20.25
20.33
20.36
20.38
20.35
20.31
20.24
20.14

20.23
20.28
20.41
20.47
20.47
20.34
20.22
20.10

20.25
20.27
20.20
20.08

20.21
20.20
20.21
20.11

20.19
20.33
20.45
20.52
20.59
20.54
20.46
20.40
20.30

20.68
20.76
20.74
20.63
20.47
20.27
20.08
20.07

20.73
20.85
20.83
20.70
20.50
20.26
20.05

20.59
20.73
20.79
20.75
20.63
20.45
20.26
20.09

0.20
Detrended Prices and Output (PAT for Both)

0.09

20.08 20.55
20.17 20.61
0.24 20.58
20.28 20.49
20.32 20.35
20.32
0.17
20.29
0
20.28
0.14
0.03 20.26
0.26

20.49
20.48
20.48
20.44
20.38
20.30
20.21
0.13 20.14
0.27 20.06

20.74
20.86
20.85
20.72
20.53
20.30
20.08

20.51
20.62
20.68
20.65
20.55
20.39
0.04 20.23
0.14 20.07

20.60
20.59
20.58
20.54
20.49
20.41
20.31
20.19
0.13 20.07

20.55
20.72
20.81
20.80
20.69
20.50
20.32
20.15

20.51
20.49
20.46
20.39
20.32
20.21
20.09

0.01

0.05
0.19

20.51 20.26 20.44
20.59 20.15 20.28
20.44
0.06
0.02
20.29
0.25
0.26
0.10 20.21
0.37
0.40
0.21 20.04
0.45
0.50

0.15 20.37 20.54
0.18 20.26 20.39
0.20 20.09 20.19
0.29
0.09
0.03
0.35
0.27
0.25
0.38
0.40
0.42
0.41
0.42
0.41
0.39
0.41
0.40
0.40
0.41
0.38

0.10
0.17
0.23
0.33
0.36
0.41
0.48
0.52
0.54

4
20.04 20.16
0.32 20.25 20.23 20.51 20.16 20.40
3
20.02 20.02 20.40 20.25 20.22 20.58 20.05 20.24
2
0.01 20.09 20.31 20.17 20.22 20.45
0.14
0.06
1
0.10 20.05 20.14 20.05 20.09 20.30
0.33
0.31
0
0.06 20.09 20.17
0.06
0.13 20.24
0.45
0.45
21
0.24
0.14
0
0.17
0.25 20.08
0.53
0.56
22
0.27
0.15 20.02
0.20
0.21 20.02
0.52
0.52
23
0.26
0.18
0.02
0.32
0.25
0
0.45
0.43
24
0.23 20.03
0.05
0.42
0.26
0.10
0.40
0.37
Note: Shaded cells indicate instances in which the difference is significant at the 10% level.

0.28 20.29 20.50
0.32 20.18 20.36
0.36 20.02 20.16
0.45
0.16
0.07
0.51
0.32
0.29
0.56
0.46
0.46
0.59
0.47
0.45
0.56
0.46
0.44
0.55
0.46
0.42

0.22
0.30
0.37
0.47
0.50
0.56
0.61
0.64
0.64

0.01

0.18
0.09
0.21
0.10
0.21
0.13
0.17 20.06

0.03
0.01
0.06
0.09

0.09
0.22
Inflation and Detrended Output (HP Filter)

20.34
20.30
20.26
20.21
20.14
20.05

20.57
20.77
20.86
20.83
20.69
20.46
20.26
20.06

0.03
0.14
0.18
0.17
0.04
0.45
0.47
0.31
0.20
0.04
0.37
0.39
0.41
0.21
0.15
0.32
0.32
Inflation and Detrended Output (PAT)

FEDERAL RESERVE BANK

80

OF

S T. L O U I S

0.24

0.08

S E P T E M B E R /O C T O B E R 1996

The differences between the correlations across the assigned break point that
are significant at a 10 percent level are
shaded in Table 5.5 In the case of Taiwan,
33 of 72 reported correlations are
significantly different. This result is consistent with the findings of Backus and Kehoe
(1992), who report that the output-price
correlations vary across countries and time
periods. It is also consistent with Gavin
and Kydland (1995), who find a significant
break in the covariance structure of output
and prices for the United States, which
they attribute to a policy change in 1979.
In the case of Korea, only five of the
seventy-two differences are significant at
the 10 percent level. The lower rejection
rate for Korea is associated with both fewer
large differences and the smaller sample
size used to construct the test statistics.

factor in output fluctuations, while Jun
(1992) finds that aggregate demand shocks
are important. These two researchers both
used a slightly different structural VAR
model of the Blanchard and Quah (1989)
type.
Further research is needed to reconcile
these contradictory findings. For instance,
in cases where covariance structure is stationary, it may be useful to apply a coherence measure in spectral analysis,
which represents the proportion of the
variance of either series that can be explained by the other, frequency by frequency. In cases where the correlation
structure between output and prices has
changed, it is important to use theories for
a better understanding of why the changes
have occurred.

CONCLUSION

Alba, Joseph P., and Papell, David H. “Trend Breaks and the Unit Root
Hypothesis for Newly Industrializing and Newly Exporting Countries,”
Review of International Economics (October 1995), pp. 264–74.

REFERENCES

The main finding of this paper is that,
in Korea and Taiwan, the detrended price
level is negatively correlated with cyclical
output, while inflation is positively correlated with the cyclical component of output. The results generally hold, whether
price is measured by the deflator or by the
consumer price index, for a number of filtering procedures and for subsamples
before and after the estimated dates of
structural breaks in output. This result
confirms the findings of Chadha and
Prasad (1994) in G-7 countries, that the
price level is countercyclical, while
inflation is procyclical.
The results do not seem to suggest any
decisive conclusions regarding the appropriate model for the business cycle. While
the countercyclical behavior of the price
level is consistent with the predictions the
supply-determined models of the business
cycle, the procyclical behavior of the inflation rate is consistent with the predictions
of conventional demand-determined
models of the business cycle. This finding
is consistent with the recent contradictory
findings concerning the sources of economic fluctuations in Korea; Yoo (1992)
finds that supply shocks are a dominant

Backus, David K., and Patrick J. Kehoe. “International Evidence of the
Historical Properties of Business Cycles,” The American Economic
Review (September 1992), pp. 864–88.
Banerjee, Anindya, Robin L. Lumsdaine, and James H. Stock. “Recursive
and Sequential Tests of the Unit Root and Trend-Break Hypotheses:
Theory and International Evidence,” Journal of Business and Economic
Statistics (July 1992), pp. 271–87.
Beveridge, Stephen, and Charles R. Nelson. “A New Approach to
Decomposition of Economic Time Series into Permanent and Transitory
Components with Particular Attention to the Measurement of the
‘Business Cycle,’” Journal of Monetary Economics (March 1981),
pp. 151–74.
Blanchard, Olivier Jean, and Danny Quah. “The Dynamic Effects of
Aggregate Demand and Supply Disturbances,” The American Economic
Review (September 1989), pp. 655–73.
Bry, Gerhard, and Charlotte Boschan. Cyclical Analysis of Time Series:
Selected Procedures and Computer Programs, National Bureau of
Economic Research, 1971.
Chadha, Bankim, and Eswar Prasad. “Interpreting the Cyclical Behavior
of Prices,” International Monetary Fund Staff Papers (June 1993),
pp. 266–98.
_______. “Are Prices Countercyclical? Evidence from the G-7,”
Journal of Monetary Economics (October 1994), pp. 239–47.
Christiano, Lawrence J. “Searching for a Break in GNP,” Journal of
Business and Economic Statistics (July 1992), pp. 237–50.

FEDERAL RESERVE BANK

81

OF

S T. L O U I S

S E P T E M B E R /O C T O B E R 1996

Serletis, Apostolos. “International Evidence on Breaking Trend Functions
in Macroeconomic Variables,” Applied Economics (February 1994),
pp. 175–79.

Cogley, Timothy, and James M. Nason. “Effects of the Hodrick-Prescott
Filter on Trend and Difference Stationary Time Series: Implications for
Business Cycle Research,” Journal of Economic Dynamics and Control,
(Jan.-Feb. 1995), pp. 253–78.

Smith, R. Todd. “The Cyclical Behavior of Prices,” Journal of Money,
Credit, and Banking (November 1992), pp. 413–30.

Cooley, Thomas F., and Lee E. Ohanian. “The Cyclical Behavior of
Prices,” Journal of Monetary Economics (August 1991), pp.
25–60.

Yoo, B. Sam. “Effects of Demand and Supply Shocks on the Korean
Economy,” Monetary and Financial Studies Working Paper #52, Bank
of Korea Institute for Monetary and Economic Research, 1992.

Gavin, William T., and Finn E. Kydland. “Endogenous Money Supply and
the Business Cycle,” Working Paper 95-010, Federal Reserve Bank of
St. Louis, 1995.

Zarnowitz, V.“Business Cycles: Theory, History, Indicators, and
Forecasting,” The University of Chicago Press, 1992.

Gregory, Allan W. and Gregor W. Smith. “Business Cycle Theory and
Econometrics,” The Economic Journal (November 1995), pp.
1597–1608.

Zivot, Eric, and Donald W. K. Andrews. “Further Evidence on the Great
Crash, the Oil Price Shock and the Unit Root Hypothesis,” Journal of
Business and Economic Statistics (July 1992), pp. 251–70.

Hall, Thomas E. “Price Cyclicality in the Natural Rate-Nominal Demand
Shock Model,” Journal of Macroeconomics (Spring 1995),
pp. 257–72.
Harvey, A.H. and A. Jaeger. “Detrending, Stylized Facts and the Business
Cycle,” Journal of Applied Econometrics (July-Sept. 1993),
pp. 231–47.
Hodrick, Robert J. and Edward C. Prescott. “Post-War U.S. Business
Cycles: An Empirical Investigation,” Discussion Paper 451,
Northwestern University, Revised May 1981.
Judd, John P. and Bharat Trehan. “The Cyclical Behavior of Prices:
Interpreting the Evidence,” Journal of Money, Credit, and Banking
(August 1995), pp. 789–97.
Jun, S. “Money, Prices, Nominal Wages: Long-Run and Short-Run
Dynamics,” Korea Development Institute Studies (Spring 1992),
pp. 37–60.
King, Robert G. and Sergio T. Rebelo. “Low Frequency Filtering and Real
Business Cycles,” Journal of Economic Dynamics and Control (Jan.Mar. 1993), pp. 207–31.
Klein, Philip A. and Geoffrey Hoyt Moore. “Monitoring Growth Cycles in
Market-Oriented Countries: Developing and Using International
Economic Indicators,” Ballinger, 1985.
Kydland, Finn E. and Edward C. Prescott. “Time to Build and Aggregate
Fluctuations,” Econometrica (November 1982), pp. 1345–70.
_______. “Business Cycles: Real Facts and a Monetary Myth,”
Quarterly Review, Federal Reserve Bank of Minneapolis (Spring
1990), pp. 3–18.
Nelson, Charles R. and Charles I. Plosser. “Trends and Random Walks in
Macroeconomic Time Series: Some Evidence and Implications,“
Journal of Monetary Economics (September 1982), pp. 139–62.
Ostle, Bernard. Statistics in Research. Ames, Iowa: Iowa University
Press, 1963.
Perron, Pierre. “The Great Crash, the Oil Price Shock, and the Unit Root
Hypothesis,” Econometrica (November 1989), pp. 1361–401.

FEDERAL RESERVE BANK

82

OF

S T. L O U I S