Full text of Review (Federal Reserve Bank of St. Louis) : September/October 2007, Vol. 89, No. 5

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

Currency Design in the United States and Abroad:
Counterfeit Deterrence and Visual Accessibility
Marcela M. Williams and Richard G. Anderson
Despite the increasing use of electronic payments, currency retains an important role in the payment system of every country. In this article, the authors compare and contrast trade-offs among
currency design features, including those primarily intended to deter counterfeiting and those to
improve usability by the visually impaired. The authors conclude that periodic changes in the
design of currency are an important aspect of counterfeit deterrence and that currency designers
worldwide generally have been successful in efforts to deter counterfeiting. At the same time,
currency designers have sought to be sensitive to the needs of the visually impaired. Although
trade-offs among goals sometimes have forced compromises, new technologies promise banknotes
that are both more difficult to counterfeit and more accessible to the visually impaired. Among
the world’s currencies, U.S. banknotes are the notes most widely used outside their country of
issue and thus require special consideration. (JEL E42, E51)
Federal Reserve Bank of St. Louis Review, September/October 2007, 89(5), pp. 371-414.

An efficient and sustainable payment system development process ought to be built around the current and
foreseeable payment needs of the users in the economy,
such as consumers, government and financial and nonfinancial businesses. These needs should be weighed
against the current economic and technical capabilities
of the economy to supply the required payment instruments and services in a cost-effective manner.
Guideline 4, Committee on Payment and Settlement
Systems, Bank for International Settlements (2006)

D

espite the increasing popularity of electronic payments, currency remains an
integral part of all nations’ payment
systems and the most familiar medium for daily
face-to-face payments. In most nations, the bulk
of the circulating currency is that issued by that
nation’s monetary authority; in a few cases, the

currency of a foreign nation, such as the United
States, is widely used. Regardless of the specific
currencies in circulation, each nation’s payment system has a large number of participants,
including consumers, retail businesses, financial
institutions, the central bank or other monetary
authority, armored car companies and others who
transport currency, and third-party currency
processors. Everyone who touches currency in
a nation’s payment system has a stake in recognizing the authenticity of currency (“authenticating”), readily determining its denomination
(“denominating”), and—subject to these constraints—minimizing the cost of its handling.
To sustain public confidence in a circulating
currency, banknotes must contain features that
the public can use to readily judge their authenticity and determine the notes’ denominations.

Marcela M. Williams is a senior research associate and Richard G. Anderson is a vice president and economist at the Federal Reserve Bank
of St. Louis. Views expressed herein are solely those of the authors and are not necessarily those of the Federal Reserve Bank of St. Louis, the
Board of Governors of the Federal Reserve System, the U.S. Treasury, or the Department of Homeland Security. The authors thank numerous
colleagues for comments. They also thank, for their cooperation and assistance, the Reserve Bank of Australia, the Bank of Canada, the Bank
of Japan, the Sveriges Riksbank, the European Central Bank, the Monetary Authority of Hong Kong, the Hongkong and Shanghai Banking
Corporation Limited, the Monetary Authority of Singapore, the Swiss National Bank, the Bank of England, and the United States Bureau of
Engraving and Printing. All currency images used within are the property of the respective currency issuers and are used with permission.
Responsibility for errors remains with the authors.

© 2007, The Federal Reserve Bank of St. Louis. Articles may be reprinted, reproduced, published, distributed, displayed, and transmitted in
their entirety if copyright notice, author name(s), and full citation are included. Abstracts, synopses, and other derivative works may be made
only with prior written permission of the Federal Reserve Bank of St. Louis.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

371

Williams and Anderson

Currency design is driven by these considerations,
subject to constraints imposed by the desire to
minimize the notes’ cost and maximize their ease
of handling by third parties. Monetary authorities,
quite naturally, prefer banknotes that are relatively
inexpensive to produce and, other things equal,
have a long life in circulation. These considerations make banknote design a mixture of art,
science, and economics.1 Generally, monetary
authorities worldwide recognize that improved
public understanding and awareness of the features of their currency assist them in meeting
these goals. Yet, studies of public awareness of
currency features suggest that often many members of the public are not aware of even a single
security feature in their nation’s banknotes.2
In this article, we compare and contrast design
features in banknotes issued by monetary authorities in a number of countries, including the
United States. Our focus is on two primary design
issues: counterfeit deterrence and high visual
accessibility (for both the normal-sighted and
visually impaired).3 We discuss both those fea1

For the economist reading this article, we note that the “currency
design problem” may be framed as a constrained optimization
problem in which the objective is to maximize the contribution of
circulating currency to social welfare. Absent constraints (both
economic and technical), social welfare is maximized by producing
banknotes that contain “all” desirable features, including those
that deter illegal reproduction, make denomination rapid and
accurate, make the currency easily used by the blind and visually
impaired, and (as with machine-readable features) minimize the
cost of high-volume currency handling by third-parties. Because
it is reasonable to believe that the objective functions of monetary
authorities differ (for example, some might have a relatively larger
fraction of their banknotes in circulation abroad, or might have a
relatively more highly educated population, or might face relatively
higher labor costs for currency handling), the solutions to their
optimization problems—the resulting currency designs—will differ.
The underlying structure of our article is such an “optimization
subject to constraints” framework. In our opinion, such formalism
contributes little to the public’s understanding of currency design
issues and, hence, for accessibility to the general reader we avoid
the use of such jargon in the article. Interested readers should examine Tables 1 and 2, in which the column headings approximately
correspond to the constraints that would appear in a formal KuhnTucker optimization statement of the currency design problem.

2

Published studies on this topic are relatively difficult to locate. An
interesting paper is de Heij, DiNunzio, and Strube (2003), which,
for the United States, summarizes surveys of public awareness
conducted by Deloitte & Touche for the U.S. Bureau of Engraving
and Printing. For a recent study of the euro, see de Heij (2006).

3

Throughout this manuscript, we use the term “visually impaired”
to refer to persons with limited visual capacity that affects their
ability to carry on their daily lives, and “normal-sighted” to

372

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

2007

tures that are readily visible to consumers and
businesses, and machine-readable features not
accessible without machine assistance. Our
emphasis is on design features that the public
might use to determine the authenticity and
denomination of a banknote. In this respect, our
views coincide with those expressed by currency
experts such as Brian Lang (2002), then head of
currency at the Reserve Bank of New Zealand,
who emphasized that building sophisticated
security features into banknotes is of limited
effectiveness if the public is unaware of the features, and that a highly aware public is the best
defense against the circulation of counterfeit
banknotes. Throughout, our analysis is based on
publicly available information regarding currency
design considerations and features. To the extent
that banknotes contain covert or secret features,
these are not discussed here.
U.S. currency is an especially complicated
case: Unique in the world, U.S. currency presents
special issues because of its widespread usage
outside the United States. Although U.S. currency
has been held abroad since before the Second
World War, outflows accelerated during the 1970s.
Recent estimates are that more than 50 percent of
all U.S. banknotes in circulation are held outside
the United States.4 The worldwide presence
increases its attractiveness to counterfeiters while,
at the same time, councils caution with respect to
significant design changes to deter counterfeiting
or improve visual accessibility that might confuse
the public outside the United States. In foreign
nations experiencing political or economic uncertainty, U.S. banknotes are an attractive asset;
much of the recent growth in demand for U.S.
banknotes has been in countries of the former
Soviet Union and Latin America. Indeed, anecdotal press reports tell of Moscow taxi drivers
insisting to be paid in U.S. dollars rather than
rubles. Other stories tell of merchants in the most
describe those persons who, while perhaps having some diminished vision, are able to conduct daily major life functions, albeit
in some cases with an accommodation or limitations. We recognize
that these terms are imprecise and that significant disagreement
remains regarding the use of such highly descriptive terms. Among
other sources, a more complete discussion is available, for example,
in National Research Council (1995).
4

U.S. Treasury Department (2006b).

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

remote areas of China accepting—and giving
change—in U.S. banknotes. The extensive, widespread use of U.S. banknotes benefits American
taxpayers because, unlike Treasury bonds, the
banknotes are a liability of the U.S. monetary
authorities on which no interest is paid. The use
of the banknotes also is a social and economic
benefit to the residents of foreign countries who
might otherwise lack a currency stable in value
and widely accepted in transactions. This same
popularity also encourages counterfeiting. Counterfeiters range from the casual, who produce a few
notes with desktop scanners and ink jet printers
or with color copiers, to professionals using
sophisticated lithographic printing systems, to
foreign governments that print counterfeit “supernotes” on government-owned intaglio presses.5
Below, we consider further the ways in which
security requirements and a desire for universal
acceptability around the world may limit design
options for U.S. currency.6
We begin by exploring the characteristics of
an “ideal” currency and then examine the realworld considerations that affect currency design.
We conclude with a comparative analysis of currency features around the world.

WHAT ARE THE
CHARACTERISTICS OF AN
IDEAL CURRENCY?
There are relatively few published, analytical
studies of currency design. The studies of which
we are aware focus on U.S. currency and, for the
most part, have been funded by the U.S. Treasury’s
Bureau of Engraving and Printing (BEP) and conducted by the National Research Council (NRC),
including BEP (1983) and NRC (1993, 1995, 2006,
2007). The U.S. Treasury, in collaboration with
other government agencies, has published three
studies regarding counterfeiting and the demand
for U.S. banknotes abroad (U.S. Treasury, 2000,
2003, 2006b). Our analysis necessarily draws
heavily on these studies.
5

6

The U.S. BEP was established during the U.S.
Civil War in an effort to reduce counterfeiting.
Since that time, the primary focus of U.S. banknote design has been counterfeit deterrence—
although features that would ease usage by both
the visually impaired and normal-sighted users
have not been ignored. The NRC’s Committee on
Next-Generation Currency Design (NRC, 1993)
described the features of a “perfect” currency:
• Extremely difficult to duplicate
• Easily recognized by the general public
• Durable (remains visible after considerable
wear)
• Can be machine-readable
• Easy to produce at low cost
• Acceptable to the public (aesthetically
pleasing)
• Non-toxic and non-hazardous
Table 1 compares and contrasts banknotes’ characteristics relative to those of an “ideal currency”
for selected countries that belong to the Bank for
International Settlements’ Committee on Payment
Systems and Services. With respect to counterfeit
deterrence, the committee argued that no single
feature, by itself, can furnish adequate protection
against even the casual counterfeiter because
nearly all individual features can be simulated
by a determined, well-funded counterfeiter.
Instead, the committee recommended a “systems
approach” that combined several features with
aggressive law enforcement and detection efforts.
To introduce ideas and terminology, consider
the current U.S. $50 note (see Figure 1). On the
front, the note contains security features including
a watermark, color-shifting ink and subtle shades
of various colors; on the reverse, the note contains
in the lower right corner a visual accessibility
feature, the denomination in 14 mm tall numerals.
In the final section of this article, we examine
security and visual accessibility features of currencies from several nations.
Accessibility Features
• Large numeral on reverse

Both the National Research Council (2006) and Mihm (2006)
identify North Korea as a producer of “supernotes.”

• Large portraits

Interested readers should consult U.S. Treasury (2000, 2003, 2006b).

• Subtle over-printed colors on front

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

373

Williams and Anderson

Figure 1
U.S. $50 Note: Security and Accessibility Features

Front

1
5
2
5

4

Security Features

5

1. Enlarged, off-center portrait

6

2. Subtle, over-printed colors

7

3

3. Color-shifting ink
4. Fine-line printing patterns
5. Microprinting
6. Watermark
7. Security thread
Accessibility Features
8. Large denomination numeral on
reverse

5

• Enlarged portrait
• Colors that vary by denomination

8
Back

NOTE: See Tables 1 and 2 and the appendix for a complete listing of security and accessibility features. See also the boxed insert on
p. 408 and Figure 11 on p. 409.
SOURCE: U.S. Bureau of Engraving and Printing; www. moneyfactory.com/newmoney.

Security Features
• Watermark
• Color-shifting inks
• Fine-line printing patterns
• Enlarged off-center portraits
• Machine-readable feature
• Security thread
• Microprinting
Although the task given to the NRC’s 1993
committee on currency design was to explore
features that deter counterfeiting, the first item
on its evaluation criteria was “visual and tactile
recognizability,” defined as “an assessment of the
ease with which a U.S. citizen…could readily
recognize the feature in normal ambient illumination unaided, or aided with a simple, inexpensive
374

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

2007

device.” It is widely recognized that the “look and
feel” of currency is important for determining
whether it is genuine, both for normal-sighted
users and the visually impaired. The committee
concluded that U.S. currency’s unique banknote
paper, combined with its raised surface created by
intaglio printing, were the essential ingredients for
notes to have the correct “feel” of U.S. currency.7
More recently, the NRC has issued two detailed
technical reports on counterfeit deterrence (NRC,
7

Security intaglio printing refers to print produced from handengraved plates with a recessed image. Thick paste inks and high
printing pressures are used to create a raised surface on the note
that cannot be reproduced by planar printing processes. Virtually
all banknotes are printed, at least in part, by intaglio methods.
Additional features, such as serial numbers and color, often are
added via letterpress or offset printing, although color intaglio is
feasible. Van Renesse (2005) and NRC (2006) provide detailed
discussions of printing techniques.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

2006 and 2007). In broad terms, the findings and
recommendations resemble the 1993 report:
Currency redesign must be an ongoing process to
remain ahead of the improving technology available to counterfeiters.
In 1995, the NRC (via its Committee on Currency Features Usable by the Visually Impaired)
issued a report focused solely on visual accessibility (NRC, 1995). This study was commissioned
by the BEP, and is, as far as we are aware, the most
extensive on the topic. We review its recommendations below. In their scope and completeness,
the NRC’s 1993 and 1995 studies complement
each other. Unfortunately, the studies are almost
completely compartmentalized in their suggestions for an ideal currency. The studies omitted
comparisons between design features to deter
counterfeiting and to increase visual accessibility:
The 1993 report (NRC, 1993) considered “ideal”
features for counterfeit deterrence, while the 1995
report (NRC, 1995) considered “ideal” visual
features. The NRC’s 1993 report, for example, in
approximately 125 pages, never mentions concerns of the blind or visually impaired.8 The NRC’s
1995 report, for example, concluded that “features
currently used by other countries…can be considered for inclusion in the forthcoming redesign
of U.S. currency without compromising the security of American banknotes” (NRC, 1995, p.11).
We discuss below features that have been introduced in newer currency designs.
Our analysis of notes issued worldwide suggests that security-related features (including
those to deter counterfeiting) and features that
enhance visual accessibility often are complementary. For U.S. banknotes, designing currency
robust to counterfeiters but also as highly visually
accessible as possible perhaps is more challenging
due to the widespread use of U.S. dollars abroad;
yet, the Treasury’s successful education campaign
that surrounded introduction of the redesigned
$100 note suggests that education can overcome
initial resistance to design changes.9
8

The 1993 report is the first report specifically linked to a pending
currency redesign. Two previous reports by the NRC are discussed
in the 1993 document. One of those earlier reports, in 1985, recommended the security thread and small-size intaglio-printed characters implemented in Series 1990 U.S. banknotes.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

CURRENCY DESIGN FEATURES
TO DETER COUNTERFEITING
Technological innovation in color copying,
scanning, and printing is the most important factor propelling increasingly sophisticated currency
designs to deter counterfeiting.
Innovation has intensified the race between
increasingly sophisticated banknote counterfeiters
and banknote designers. In many lines of business,
the improving performance and decreasing cost
of information technology have reduced “barriers
to entry”—unfortunately, perhaps, the same is
true in counterfeiting. The NRC’s most recent
technical report (NRC, 2007) concludes that this
race eventually will be lost for the current generation of paper-based banknotes and explores a
wide range of innovative substrates as a replacement for current banknote paper. The report’s
overall conclusion echoes the conclusions of
previous NRC reports (NRC, 1993, 2006) that only
continuous design innovation can sustain a low
frequency of counterfeit notes in circulation. The
report’s engineering analysis of alternative substrates is fascinating but speculative. We leave
details of the report to the interested reader and
focus herein on current banknote designs and
features.
Banknote designers worldwide have introduced a number of features to deter counterfeiting. Features to assist authentication for selected
countries are summarized in the first column of
Table 1. Such features include novel substrates
such as polymer with clear windows that include
embossed and printed security features10; embedded security threads, including ones that glow
different colors under ultraviolet (UV) light11;
high-relief raised intaglio printing; extra color
9

See U.S. Treasury (2000, 2003, and 2006b) for discussions of
uncertainty generated abroad by design changes and the accompanying education campaigns. Lambert and Stanton (2001) discuss
holdings of U.S. notes abroad relative to the Series 1996 and 2004
design changes.

10

The Reserve Bank of Australia issued the world’s first polymer
note series between 1992 and 1996. The NRC (2006, 2007) reviews
the engineering details of such plastics and provides a list of the
numerous countries now using such currency.

11

Among others, U.S. banknotes now contain such threads.

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

2007

375

Williams and Anderson

Table 1
Currency Features for Selected Countries, As Reported by the Currency Issuers
Features to assist authentication
and denomination
(anti-counterfeiting measures)

Features to assist the visually impaired
(visual features)

Australia

Clear window
Made of polymer substrate instead of paper
See-through registration device
Shadow image
Intaglio printing
Background print (offset)
Microprinting
Fluorescent ink properties

Different sizes
Different colors
Intaglio printing
Large numerals

Canada

Holographic stripe
Watermark portrait
Windowed color-shifting thread
See-through number
Raised ink/intaglio printing
Fine-line patterns
Microprinting
Fluorescence/UV-light tests
Serial number
Colors

Different colors
Tactile symbols
Large high-contrast numerals
Machine readable

“Feel, look and tilt”
Made of pure cotton
Raised print/intaglio printing
Tactile marks
Watermark
Security thread
See-through number
Perforations
Hologram patch (€50 €100, €200, €500)
Color-changing number (€50, €100, €200, and €500)
Glossy/iridescent stripe
Microprinting
UV-light tests

Different sizes
Striking colors
Large numerals
Intaglio printing
Tactile marks (€200 and €500)

Euro Currency Area*

Hong Kong SAR

Fluorescent machine-readable barcode
Iridescent images
Denomination numeral in optical variable ink
Holographic windowed thread
Concealed image/denomination
Intaglio printing
Fluorescent fibers/UV-light tests
Watermarks
Security thread
See-through feature
Serial numbers

Different sizes
Different colors
Large numerals
Intaglio printing

NOTE: *For the 12 member states of the European Monetary Union participating in the single currency. †Life span of euro banknotes is
calculated by dividing, for each denomination, the average circulation of banknotes during the previous 12 months by the number of
banknotes sorted to unfit during the previous 12 months.
SOURCE: European Central Bank.

376

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

Durability

Cost of production

Bank of Australia estimates that polymer notes
last at least four times longer than paper notes

Costs are confidential

C$5: 1-2 years
C$10: 1-2 years
C$20: 2-4 years
C$50: 4-6 years
C$100: 7-9 years

Approximately CA$0.09 per note

€5: 1.2 years†
€10: 1.4 years
€20: 1.5 years
€50: 3.7 years
€100: 8 years
€200: 11.8 years
€500: 25.6 years

€0.08 per banknote

Approximately 2 years

Costs are confidential

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

377

Williams and Anderson

Table 1, cont’d
Currency Features for Selected Countries, As Reported by the Currency Issuers
Features to assist authentication
and denomination
(anti-counterfeiting measures)
Japan

Singapore

Sweden

Features to assist the visually impaired
(visual features)

Watermark
Ultra-fine-line printing
Intaglio printing
Microprinting
Luminescent ink
Hologram
Latent image
Pearl ink

Tactile marks/intaglio printing
Different sizes
Different colors

Microprinting
Kinegram
Lithographic print
Engraved portrait
Braille codes
Asymmetrical serial numbers (vertical and horizontal)
Anti-color-copying line structures
Lift twin
Intaglio printing
Highlight watermark
Watermark
Perfect registration
Security thread
Invisible feature: non-reflective under UV light
See-through windows on polymer notes

Braille codes
Different colors
Different sizes
Intaglio printing

Watermark
Security thread
See-through picture
Intaglio printing
Additional features for banknotes with foil strip:
Fluorescent picture
Microlettering
Foil strip–hologram
Shimmering mother-of-pearl ink
Motion feature

Different colors
Different sizes
Intaglio printing

Transparent register
Watermark digits
Intaglio printing (digits)
Perforated number
Optically variable ink
UV digits
Metallic digits
Tilt effect

Different colors
Different sizes
Perforated number

Switzerland

NOTE: ‡Costs of producing a banknote (conception and design, paper, printing, information) average approximately 30 centimes per
new note. Yearly production assuming a 3-year lifespan is 10 centimes per note in circulation.

378

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

Durability

Cost of production

¥1,000: 1-2 years
¥5,000: 1-2 years
¥10,000: 4-5 years

¥15.1 per banknote

S$2 polymer: 72-96 months
(issued in Jan 2006, still monitoring the performance)
S$5 paper: 18-24 months
S$10 paper: 24-36 months
S$10 polymer: 72-96 months
(issued in May 2004, still monitoring the performance)
S$50 paper: 24-36 months
S$100 paper: 48-60 months
S$1,000 paper: 48-60 months
S$10,000 paper: 48-60 months

S$2 polymer: S$0.10
S$5 paper: S$0.10
S$10 paper:S$0.09
S$10 polymer: S$0.12
S$50 paper S$0.08
S$100 paper S$0.07
S$1,000 paper S$0.09
S$10,000 paper S$0.55

Varies by denomination
20 kronor: about 1 year
1,000 kronor: about 5 years

Approximately 7.5 US cents
per banknote

10 CHF: 2 years
20 CHF: 1.5 years
50 CHF: 1.5 years
100 CHF: 3.5 years
200 CHF: 3 years
1,000 CHF: 10 years

About CHF 0.30 centimes per note‡

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

379

Williams and Anderson

Table 1, cont’d
Currency Features for Selected Countries, As Reported by the Currency Issuers
Features to assist authentication
and denomination
(anti-counterfeiting measures)

Features to assist the visually impaired
(visual features)

Cotton and linen paper
Metallic thread
Watermark
Print quality
Hologram
UV-light tests
Microlettering
Foil (on £50 note)

Different sizes
Large denomination numerals
Densely colored recognition symbols
to indicate denomination:
£5: turquoise circle,
£10: orange diamond,
£20: purple square (on the old £20 note),
£50: red triangle

Watermark
Color-shifting inks
Fine-line printing patterns
Enlarged off-center portraits
Low-vision feature
Machine-readable feature
Security thread
Microprinting

Oversize denomination numeral on reverse
Larger portraits (than previous designs)
Subtle over-printed colors on front

United Kingdom

United States

NOTE: §In the year to end-February 2006, 978 million notes were produced at a total cost of £28 million. ¶Figure based on 2001 fiscal
year figures: A total of 7,004,800,000 notes were produced, and $219,240,000 was expended on currency production.

features on bills; magnetic ink; and various
embedded machine-readable features.12 Counterfeit deterrence features are discussed further in
the final section of this article.
United States banknotes contain a large number of security features to deter counterfeiting,
including embedded security threads that glow
different colors by denomination under UV light;
large, complex line-drawn presidential portraits;
watermarks; embedded fibers; color-shifting ink
(the color of which differs by denomination);
microprinting too small to be reproduced by
current-generation photographic copiers and
consumer-level scanners; fine-line printing; color
over-printing in subtle shades; and embedded
machine-readable features. Machine-readable
12

Electronic handheld note readers, of significant value to the visually impaired, perhaps can use machine-readable features—but
specific technical designs are proprietary and confidential.
Although several currency issuers have informed us that the
machine-readable features of their banknotes might be suitable
for handheld denominators, to our knowledge such machines are
available only for Canadian and U.S. banknotes.

380

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

2007

features, most of which vary by denomination,
include the notes’ optical spectrum and image
(including color-shifting inks, and a fluorescent
strip that glows in denomination-specific colors
under certain frequencies of light), magnetic ink,
UV spectrum, and infrared ink pattern (NRC,
2006, pp. 12-13). The NRC notes that machinereadable features are useful both to authenticate
notes and to determine notes denomination.
Single-note denominators such as those installed
in retail vending machines, for example, “typically use infrared, broad-wavelength optical, or
magnetic sensors to detect denomination-specific
features,” while single-note authenticators “also
detect ultraviolet and fluorescent patterns.” (NRC,
2006, p. 12). The report notes that low-quality
counterfeit notes most commonly are identified
by low optical image quality, lack of magnetic
and/or infrared ink, or incorrect paper fluorescence (discussed further below). Detection of
high-quality counterfeits may require careful
sensing of magnetic, infrared, or UV signatures.
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

Durability

Cost of production

Ranges from around 1 year for the £5 note,
to 5 or more years for the £50 note

Approximately £0.029 per banknote§

$ 1: 21 months
$ 5: 16 months
$ 10: 18 months
$ 20: 24 months
$ 50: 55 months
$100: 89 months

Approximately 3 cents per banknote¶

Magnetic ink is a difficult feature for counterfeiters to replicate, although careful procedures
perhaps can maintain magnetic material in suspension long enough for printing; the NRC report
notes that 90 percent of counterfeit notes found
by high-speed verifiers are detected as a result of
incorrect magnetic signatures (NRC, 2006, p. 12).13
The paper substrate of U.S. banknotes has a
difficult-to-duplicate light green-yellow tint that
is a major deterrent to counterfeiting (NRC, 2006).
The paper is manufactured under close security
by a single U.S. firm from a mixture of 75 percent
cotton and 25 percent flax.14 The color is a characteristic of the paper manufacturing process
during which no clay or other whitening agents
are added and no bleach or other chemicals are
13

We are not aware of comparable discussions of machine-readable
features for banknotes issued by monetary authorities outside the
United States, although many authorities include machine-readable features in descriptions of their banknotes.

14

The flax provides a “stiffness” to the notes that is absent in many
100 percent cotton notes and increases their durability.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

used to whiten the paper. As a result, unlike most
printing papers, U.S. banknote paper does not
fluoresce under UV light. When combined with
raised texture provided by intaglio-printed images
and numerals, the unique “feel” of U.S. banknotes is reported to be the most common method
of counterfeit detection by the public and bank
employees. For the visually impaired, this same
feel is a tactile clue that helps to determine
whether an offered banknote is genuine or not;
it does not, however, assist the user with identifying the denomination of the note.
To foreshadow somewhat our discussion
below of visual accessibility features, we note
that the light green-yellow tint of U.S. banknote
paper limits the contrast ratio that can be achieved
between printed denomination numerals and
their background. A higher contrast ratio improves
the visual clarity of numerals (ideally, near black
on white). But achieving higher contrast likely
would require introducing bleach or other whiteners into the papermaking process. To the extent
S E P T E M B E R / O C TO B E R

2007

381

Williams and Anderson

that such a change would alter the color and “feel”
of U.S. banknote paper, it would remove one of
the primary features used by the public, including
the visually impaired, to detect counterfeit notes.
To the extent the modified paper would display
fluorescence, the change would also remove a
test for the legitimacy of the banknote paper. At
least with respect to banknote paper, anti-counterfeiting and visual accessibility considerations
interact in a complex way, illustrating the potential trade-offs in currency design.
Preserving the traditional “look and feel” of
U.S. banknotes has been an important consideration in recent design changes, and security features
have been selected so as not to change the traditional look and feel of U.S. banknotes. The NRC
lauds current security features as highly durable,
low cost, odorless, and environmentally sound
(that is, they do not depend on hazardous materials being included in the note) (NRC, 2006, p. 2).
Further, many of the features are detectable by
the naked eye.
All monetary authorities, on a continuing
basis, monitor the effectiveness of current counterfeit deterrence features and seek to develop new
ones. By all estimates, banknote design worldwide
has been successful in deterring counterfeiting.15
In the United States, this work is coordinated by
the Advanced Counterfeiting Deterrence Steering
Committee, which lists effective currency design
first among the three elements that comprise its
counterfeit deterrence program (U.S. Treasury,
2006b, p. 7).16 Discussions that seek to measure
counterfeiting must separate two concepts: (i) the
fraction of outstanding banknotes that are counterfeit; and (ii) for the volume of notes that flow
through a central bank’s cash offices, what fraction
are determined to be counterfeit. Counterfeit notes
may be detected at any point in the economy,
15

The appendix contains a list of internet sources for counterfeiting
volume statistics of major world currencies.

16

The second and third, respectively, are law enforcement and
public education programs. The Advanced Counterfeit Deterrence
Steering Committee is an interagency group that includes representatives from the Department of the Treasury (including the BEP),
the Federal Reserve System, and the Department of Homeland
Security’s Secret Service. (Prior to creation of the Department of
Homeland Security, the Secret Service was an agency of the
Treasury Department.)

382

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

2007

including by retail merchants, bank tellers, or
central bank cash offices. Although central bank
processing centers typically own the most sophisticated equipment, in the United States the
majority of notes are detected by retail and bank
personnel; again, the NRC notes that an incorrect
“look and feel” is the most important detection
method (NRC, 2006).
The flow of notes through central bank cash
offices—relative to the outstanding number of
notes in circulation—differs by denomination,
by country, and through time, making it difficult
to accurately infer the fraction of counterfeit notes
in circulation. For almost all countries, the measured frequency of counterfeit notes is regarded
as “low.” For U.S. banknotes in circulation both
inside the United States and abroad, the Treasury
estimates that fewer than 1 in 10,000 are counterfeit.17 The Federal Reserve reports that it removed
from circulation during 2005, and sent to the
Secret Service, 6.4 notes per million processed,
or a total of approximately 234,000 notes among
the 36.5 billion pieces of currency processed.18
The incidence for the euro also is low: For 2005,
the European Central Bank (ECB) reports detecting and confiscating from circulation 579,000
notes during the annual processing of approximately 30 billion notes (as of December 31, 2005,
there were 10.4 billion euro notes in circulation).19
The €50 and €100 were the most frequently
counterfeited notes.
17

The Secret Service has jurisdiction over all matters related to
counterfeit Federal Reserve notes. For discussion of methodologies
used to estimate counterfeit currency worldwide, see the interagency reports issued by the U.S. Department of the Treasury
(2000, 2003, and 2006b).

18

During 2005, the Federal Reserve Banks processed 36.5 billion
pieces of currency (Board of Governors of the Federal Reserve
System, 2006). For the number of detected counterfeits, see U.S.
Department of the Treasury (2006b, Table 6.3). As of December 31,
2005, 24.9 billion banknotes were in circulation, excluding $2
notes (U.S. Treasury, 2006a). Of these, approximately 5.4 billion
were $100 notes, a large proportion of which are likely held abroad,
and approximately 8.8 billion were $1 notes, likely seldom
counterfeited.

19

See ECB (2006a, p. 102) and the biannual report of the ECB
regarding counterfeiting, available at www.ecb.int/press/pr/date/
2006/html/ pr060113_1.en.html. European newspaper articles suggest that counterfeiting decreased sharply following introduction
of the euro, an example of the familiar result that design changes
tend to slow counterfeiting. See also europa.eu/scadplus/leg/en/
s30003.htm, a website of the European Parliament, and Europol
(2006a, 2006b)

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

Changes in currency design are an important
tool to deter counterfeiting. European newspaper
articles suggest that counterfeiting rates for the
euro are significantly lower than for the individual
country banknotes that it replaced. News reports
from Canada also illustrate the necessity for updating currency designs. Published reports for 2004
cited the counterfeiting of older-design notes as
being at “dangerous” levels, with a rate of 4.7
counterfeit notes per 10,000 notes examined.20
Published reports affirm that the new-design
notes with enhanced deterrence features sharply
reduced the rate to 2.2 notes per 10,000 during the
first eight months of 2006; features of Canadian
banknotes are examined below. The Reserve Bank
of New Zealand and the Reserve Bank of Australia
reported to us that counterfeiting of polymer notes
has been negligible, with ongoing activity limited
to remaining paper notes in circulation; polymer
notes are described in more detail below.

CURRENCY DESIGN FEATURES
FOR THE VISUALLY IMPAIRED
Banknote designers worldwide are aware of
the desire by the visually impaired to use currency
in day-to-day transactions. Most banknote issuers
consult regularly with advocacy organizations for
the visually impaired. The principal issue for the
visually impaired (persons with some recognizable visual field, including approximately 90 percent of persons with visual impairment in the
United States), is to readily determine the denomination of banknotes during a transaction, say, at
a distance of 16 to 24 inches.21 The principal
issue for the blind (persons with no recognizable
visual field) is the presence of tactile banknote
20

21

More information is available from the Royal Canadian Mounted
Police (www.rcmp-grc.gc.ca/scams/counter_e.htm) and various
Canadian Broadcasting Corporation reports, including www.cbc.ca/
news/background/counterfeit/ and www.cbc.ca/canada/story/
2006/11/20/counterfeit-bills.html.
In our research, we have encountered several standards for classifying persons as “visually impaired” or “blind.” The text in this
paragraph uses one such classification. Below, we cite another in
which blind is equated to a corrected visual acuity of less than
20/200 and visually impaired as a corrected visual acuity of less
than 20/40.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

features and/or machine-readable features that
assist automated recognition of denomination by
electronic currency readers. Accessibility features
of currencies for selected countries are summarized in column 2 of Table 1.
As in most areas of economics, there is no free
lunch in currency design: Those features that are
of the highest value to the visually impaired also
may, in some circumstances, impose significant
costs on other currency users and handlers,
including retail merchants, financial institutions,
and high-volume currency processors. The application of economic benefit-cost analysis to select a
set of features is not straightforward. For example,
the benefiting group is not necessarily a closed set.
Some features that assist the visually impaired—
such as large high-contrast numerals, note sizes
that vary by denomination, and sharply different
bright primary note colors—also assist normalsighted users to denominate banknotes under lowlight conditions or during hectic transactions.
While the benefit is larger for the visually
impaired, some benefit accrues to all persons using
the notes. The benefits of features also varies by
group: Low-cost features, such as larger denomination numerals, do not benefit the blind, whereas
high-cost features, such as note sizes that vary by
denomination, benefit the blind as well as other
note users. Further, the visually impaired are not
a homogeneous group and include persons born
blind or who became blind during childhood,
persons who became blind as adults, and individuals with limited visual acuity. Some know
Braille, but many do not.
Previous studies of note design have consistently found that the design feature most valuable
to the visually impaired is note sizes that vary by
denomination.22 For example, in the summary
section of its 1983 report to the Congress, the
BEP wrote (BEP, 1983, p. 17):
22

See, for example, the BEP (1983) and the NRC (1995). Both these
studies discussed additional proposed currency design features
to assist the visually impaired, including Braille or Braille-like
tactile features and notched edges or clipped corners. We do not
discuss these features here because previous studies have concluded that the features deteriorate rapidly with the usage of the
banknote. But, see the discussion of Canadian currency below,
which uses non-Braille patterns of raised dots to assist the visually
impaired.

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

2007

383

Williams and Anderson

…[I]t would appear that the most broadly useful currency design change would be to produce Federal Reserve notes in a different size
for each denomination. However, the effects
of such a change on broad and diverse segments
of the population would be monumental.

Although the report’s authors ranked this design
feature first when measured by its assistance to
the visually impaired, they ranked it second
overall due to its cost. Implementing a system of
varying-size banknotes would affect all handlers
of banknotes, including financial institutions,
retail merchants, vending machine operators,
high-volume banknote processors, the Federal
Reserve System, the BEP, and the non-visually
impaired. Instead, the authors ranked first among
alternatives a handheld electronic note reader, in
part because of its zero impact on most participants in the currency-based segment of the U.S.
payment system (BEP, 1983, p.17).
With respect to the cost of implementing a
system of varying-size banknotes where none
exists, we are not aware of any published, public
estimates for the total cost to an economy;
undoubtedly, these costs would be dependent on
the rate of implementation relative to the normal
rate of replacement of depreciated equipment.
Some published estimates are available for the
BEP alone. In 1983, the Bureau estimated an initial
start-up cost of $25.6 million and annual incremental costs of $6.7 million (BEP, 1983, p.12). In
2006, the Bureau estimated the implementation
cost to be as high as $230 million (Andrews, 2006).
For perspective, these numbers might be compared to the cost of printing current U.S. banknotes. In 1983, the Federal Reserve System
purchased 4.6 million banknotes from the BEP,
for which the production cost (including certain
adjustments for accounting changes and equipment purchases) was $23.00 per 1,000 banknotes,
a total of $105.8 million (Volcker, 1986, Table 4).
Relative to this figure, 1983’s start-up costs equal
23 percent of the year’s printing cost; the estimated
annual incremental cost adds an additional 7
percent. If perhaps the start-up costs were to be
amortized over a decade, annual currency manufacturing costs might increase by approximately
10 percent. Consider next 2006, in which the
384

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

2007

manufacturing cost for banknotes delivered to
the Federal Reserve System was $467 million
(Board of Governors of the Federal Reserve, 2006b,
Table 1). Relative to these figures, the 2006 startupcosts are approximately 49 percent of the 2006
printing cost. If amortized over a decade, the cost
is approximately 5 percent per year, not including
any incremental annual cost.23
Advances in technology promise increased
ease of use for the blind and visually impaired.
Canadian banknotes and all U.S. new-design banknotes issued since 1997, for example, contain
specific machine-readable features that have been
designed to assist the visually impaired.24 In
Canada, through the Canadian National Institute
for the Blind, blind and visually impaired persons
23

The BEP sells banknotes to the Federal Reserve at a markup over
the cost of manufacture. Although the Federal Reserve has some
control over the quantities of notes ordered, the BEP sets the billing
rates so as to recover its costs from the Federal Reserve (and other
customers). For fiscal year 2006, BEP reports that the manufacturing
cost for notes printed averaged $27.49 per thousand (U.S. Treasury,
2007, p. 72). The BEP marks up notes (and other products), under
Public Laws 81-656 (1950) and 95-81 (1997), to cover direct and
indirect costs of operation, including administrative costs, and
amounts sufficient to fund capital investments and meet working
capital requirements (U.S. Treasury, 2007, p. 71). Denominationspecific BEP billing rates to the Federal Reserve System for notes
in fiscal year 2006 ranged from a low of $43 per thousand for $1
and $2 notes, to a high of $81 per thousand for Series 2004 notes;
for calendar year 2006, the Federal Reserve reports paying a
volume-weighted average price of $55.74 per thousand (Board of
Governors, 2006b, Table 2). The $1 and $2 notes, which have not
been redesigned, do not include the Series-1996 and Series-2004
design security features; $5 notes include watermarks, but not
color-shifting ink; Series-1996 $100 notes include watermarks
and color-shifting ink; and Series-2004 $10, $20, and $50 notes
include watermarks, color-shifting ink, background colors, and
additional security features. The percentage ratios mentioned in
the text would be smaller relative to the 2007 currency budget
(assuming the costs didn’t increase as well). For calendar year
2007, the Federal Reserve reports a currency budget of $574.3
million, an increase of $107.3 million, or 23 percent. While the
largest reason for the increase is a higher proportion of newer
design notes that are more costly to produce, almost one-third
($32.1 million) is to fund a decade-long BEP capital improvement
plan which, during the first four years, will replace production
equipment in Washington, D.C., including intaglio presses (some
up to 26 years old) and overprinting equipment for some production lines (Board of Governors, 2006b).

24

In response to our inquiries, other countries replied that their notes
contain machine-readable features that could be used to assist the
visually impaired, but that such features had not been included
specifically to assist the visually impaired. For countries other than
the United States and Canada, we are not aware of any available
handheld currency authentication or denomination devices that
use such features. The machine-readable features, of course, are
of value to other currency handlers, including financial institutions,
retail stores, and high-volume currency processors. See also
Tables 1 and 2.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

can obtain at no cost a handheld banknote reader
that signals denomination to the user via voice,
tone, or vibration. This machine is produced by
Brytech (www.brytech.com), which introduced a
reader for Canadian currency in 1989 and a reader
for U.S. currency in 1992. The readers are not
identical: The reader for Canadian notes is smaller
and contains features not included in the U.S.
reader. In our research, we purchased a Brytech
“Note Teller 2” reader for U.S. banknotes; the
reader measures 6 × 3 × 1 inches and weighs 8
ounces.25 The manufacturer does not publish
recognition accuracy statistics for the reader (but
offers a 30-day money-back guarantee). We asked
approximately 20 of our colleagues to test their
banknotes, both old and new designs, in a variety
of conditions. Since the note reader “reads” the
ends of the banknotes, it failed to read only those
notes with folded corners and/or stains at the
ends.26 Brytech staff noted that (i) although some
customers find the U.S. currency reader satisfactory, others find it bulky and too expensive; and
(ii) some users have reported problems using it
with newer 1996 series and later currency designs.
Although not inexpensive (at $280 U.S.), perhaps
technological advances will allow future price
reductions. Recently, recognition of U.S. banknotes has been added to the Kurzweil-National
Federation of the Blind reader.27 This reader,
according to published documentation, captures
and parses a digital image of an object, suggesting
it is not using machine-readable features of the
notes. We are aware of no published figures regarding accuracy of this device. Further, because it
uses only a visual scan of the note, it perhaps is
less valuable for authenticating than denominating notes. Although the currency feature will be
of value to owners of the Kurzweil reader, the
25

Our Note Teller 2 reader was the “standard” model with voice
features. A recently introduced “enhanced” model includes the
signal-by-vibration feature available on the Canadian currency
model but not previously available on the U.S. model.

26

The specific machine-readable features and technology used by
Brytech are trade secrets.

27

Details are at www.knfbreader.com. The web site says the Kurzweil
reader makes a digital image of the object, then parses the digital
image. Hence, the reader uses no machine-readable features of the
notes.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

reader’s $3,500 price very likely places it beyond
the means of many visually impaired persons.
When considering costly design features that
assist the visually impaired, an economic analysis
of the benefit-cost trade-off depends, in part, on
the number of affected persons. But, a great deal
of uncertainty surrounds such estimates: It is
difficult to measure accurately the number of
Americans who will today and might in the future
benefit from improved banknote visual accessibility. A study published in 2006 (Vitale et al.,
2006) estimates the number of people with visual
impairment to be about 14 million.28 The NRC
(1995) reports, based on one widely used classification, that as of the early 1990s approximately
9 million Americans were classified as blind or
visually impaired. The NRC (1995) report cautions
that traditional criteria may greatly understate
the true extent of impairment because the determination is made by measuring visual acuity in
laboratory settings under ideal lighting and highcontrast conditions. To read text rapidly and
accurately in practical situations such as occur
at a retail cash register often requires numerals
three to five times larger than the sizes commonplace in laboratory visual acuity tests—if used,
such a criterion would classify millions of additional persons as visually impaired. In addition,
many studies, including NRC (1995), have noted
that the design of visually accessible currency is
as much an age-related issue as a vision-related
issue. Looking forward, the aging of the world’s
population portends visual problems; many conditions related to aging reduce vision, including
macular degeneration, glaucoma, and retinitis
pigmentosa. In addition, contrast sensitivity and
pupil size tend to decrease with age: At 80 years
old, the eye often admits only one-quarter the
light admitted for a 20-year old. Combined with
other age-related factors, the result is “a severe
reduction in visual performance under adverse
lighting conditions for older people.”29 (The first
28

This figure would correspond to the number of U.S. residents aged
12 years or older with visual acuity of 20/50 or worse. The study
also found that approximately 83 percent of those people estimated
to be visually impaired would qualify for a driver’s license if they
had corrective lenses.

29

NRC (1995).

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

2007

385

Williams and Anderson

formal study of the visual accessibility of U.S.
banknotes, reviewed below, was conducted at
the behest of the U.S. House of Representatives
committee on aging.) The National Eye Institute
(2002) estimates that the number of Americans
who are blind (that is, best corrected visual acuity of 20/200 or less in the better-seeing eye) and
have low vision (best corrected visual acuity of
20/40 or less) due to age-related eye diseases will
double over the next 30 years.
Recognizing these technical shortcomings
with respect to measuring visual impairment, the
NRC (1995, p. 6) argued for an aggressive stance
on vision enhancements: Small enhancements,
adequate in laboratory settings, are likely to produce designs that are inadequate in practical settings. Among their conclusions was this:
It is clear that a major need exists for a better
means of banknote denomination for the 3.7
million Americans with visual disabilities,
with the goal of giving this population the full
access to currency handling available to the rest
of society and to visually disabled people in
other countries. In addition, due to the increasing number of older individuals with impaired
vision due to minor eye disease or the normal
aging process, such features would be of great
benefit to a far wider population than that
represented by the current statistics on blindness and low vision. Certain new features,
such as color and size, and enhanced existing
features, such as larger numerals of higher
contrast, would also benefit those with normal
vision by making denomination more rapid
and convenient for all.

Visual accessibility features of banknotes
issued by several member countries of the Bank
for International Settlements’ Committee on
Payment and Settlement Systems (including the
United States) are summarized in Table 1. Features
include varying note sizes by denomination; oversize numerals to indicate denomination; highcontrast numerals printed in dark inks against
light backgrounds, or the reverse; different principal colors for different denominations; and,
tactile features such as embossed numerals or
patterns of recessed or raised dots.
In Europe, the designers of euro notes at the
ECB “consulted extensively” on the design of the
386

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

2007

euro banknotes with the European Blind Union
and the World Blind Union, the latter a worldwide
umbrella organization for advocacy groups of the
blind and visually impaired. A number of their
specific efforts are described on the web site of
the European Blind Union (www.euroblind.org)
under the “Access to Information” heading.
Special features include (i) sizes that vary by
denomination, (ii) “striking” “clearly contrasting”
colors, (iii) denominations printed in large numerals, (iv) high-relief raised intaglio printing, (v)
and special tactile marks printed in intaglio along
the edges of €200 and €500 banknotes.30
There is a similarly long history in the United
States of interaction between currency designers
at the BEP and advocacy groups for the visually
impaired. The American Council of the Blind
reportedly has submitted suggestions regularly
to the BEP since at least 1972 (NRC, 1995). In
1983, the first (to our knowledge) formal study
of the visual accessibility of U.S. banknotes was
conducted by representatives from the American
Council of the Blind, the Federal Reserve System,
the BEP, and the U.S. Secret Service, at the request
of Congressman Edward Roybal, chair of the U.S.
House Committee on Aging. The subsequent
report is BEP (1983).31 In addition to being the
first formal study, the 1983 report is important
because its issues resurface in more recent discussions. In its introduction, the report says that
“since the early 1970s” there has been a “significant interest” in designs for U.S. currency to
assist the visually impaired—that is, since the
Netherlands introduced in January 1971 innovative banknotes with raised (embossed) tactile symbols for each denomination. The report continues
by citing unpublished 1976 and 1980 studies by
the BEP regarding methods for easing the recognition of a banknote’s denomination. The 1976
study, it says, examined the incorporation of
30

See www.ecb.int/bc/euro/banknotes/visually/html/index.en.html,
as of July 2007. The €5, €10, €20, €50, €100, €200, and €500
notes, respectively, feature the colors gray, red, blue, orange, green,
yellow, and purple.

31

This report is available from the Federal Reserve Bank of St. Louis
in its FRASER archive: http://fraser.stlouisfed.org/historicaldocs/
tresbep/download/30353/study_of_mechanisms.pdf.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

Braille markings in currency, and the 1980 study
examined “notching” and/or cutting the corners
of banknotes.32
The 1983 report’s original research is presented in three parts. First, the staff reviewed the
currencies of 54 countries (including the United
States) for characteristics that would assist denomination by the visually impaired and asked each
country to provide information on the performance of the features; five countries responded.
The report’s summary of the comments notes that
“few countries have intentionally designed currency for this purpose” and that the comments
received suggest that none are “completely satisfactory.” Second, technical staff from the BEP
evaluated “selected features” with respect to cost
and usefulness to the visually impaired. Although
details are not provided, the report says that the
staff affirmed that commonly proposed design
features—including colors that varied by denomination, large and clear numerals, note sizes that
varied by denomination, cut corners or notches,
and Braille-like tactile features—likely would
assist denomination by the visually impaired.
Costs for implementing various combinations of
these features in U.S. banknotes ranged from
$600,000 to $6 million per year. Third, the staff
explored the feasibility of an automated portable
currency reader. Noting that such a device already
was available but citing its bulk and expense,
the staff recommended that funding be sought to
develop a handheld electronic currency reader
able to denominate currency. Such a device would
assist both the visually impaired and the blind,
while requiring at most only modest design
changes for U.S. banknotes.
Chronologically, the next formal U.S. study
of visual accessibility was commissioned by the
BEP to the NRC of the National Academy of
Sciences (NRC, 1995). The NRC was charged
with several tasks:
• assess features that could be used by people
who are visually disabled to recognize,
denominate, and authenticate banknotes;
32

The outcome of such experiments in not noted in BEP (1983). With
respect to notches and cut corners, the report does note that the
Federal Reserve and U.S. Secret Service “expressed fundamental
concerns” that problems might arise during a transition period.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

• recommend features that could reasonably
be incorporated into banknotes using available technology;
• suggest strategies that should be instituted
to make the recommended features more
effective;
• and identify research needs in particularly
promising areas that could lead to attractive
future approaches.
The 1995 NRC study is lengthy and complex,
considering in detail both practical and technical
issues. This quote summarizes the NRC’s view of
the state of U.S. currency design in 1995 (NRC,
1995, Chap. 2, p. 13):
The United States is alone in issuing paper
currency bills in which all denominations are
of identical size and color. The numbers on the
corners of the bills are small and of low contrast, making them difficult to read by people
with impaired vision. For everyday transactions, U.S. paper currency possesses no nonvisual identifying features, rendering it
impossible for blind people to denominate
bills without assistance. The lack of distinctive visual features and the total absence of
nonvisual features for the common user constitute a hindrance to commerce and daily
living for millions of visually disabled people.
In addition, the lack of distinctive features
results in problems of denomination for a much
wider population with mild visual impairments, including those impairments acquired
during the normal aging process, and for anyone in a poorly lit environment.

The study eventually recommends four features technologically available in 1995:
• banknote sizes that vary by denomination;
• large, high-contrast numerals on a uniform
background (for large numerals, the
Committee recommended numerals larger
than half the note’s height of approximately
65 mm);
• different predominant colors for each of
the six denominations being printed;
• and embedded features to assist development of effective, low-cost devices for
examining banknotes.
S E P T E M B E R / O C TO B E R

2007

387

Williams and Anderson

These recommendations are largely the same as
those identified in BEP (1983). Consistent with
that study, the NRC (1995) concludes that varying
banknote sizes by denomination is the most effective feature to assist denomination by both the
blind and visually impaired. The study notes,
however, that this is the most costly feature to
implement. Further, the change would primarily
benefit the blind, who are approximately 10 percent of all visually impaired persons in the United
States. In contrast, other features—such as larger
numerals and different primary colors—would
benefit all banknote users by speeding identification of denomination, particularly in low-light situations. Regarding cost, the committee concluded
that the costs to redesign U.S. banknotes and
modify equipment to handle them (especially in
varying sizes) would be high—but the Committee
did not temper its recommendations with regard
to these costs because a major redesign already
was planned during 1996-2001. Instead, the
Committee recommended that these features be
incorporated during the 1996-2001 currency
redesign cycle.
Since 1995, the BEP has introduced two new
designs for U.S. banknotes, each containing features to improve accessibility for the visually
impaired. These designs are frequently referred
to as the Series 1996 “New Currency Design” and
Series 2004 “New Color of Money” designs.33
Series 1996–design notes were introduced beginning with the $100 note in March 1996, followed
by the $50 note in October 1997, the $20 note in
the fall of 1998, and the $5 and $10 notes in May
2000. Series 2004–design notes were introduced
with the $20 note in October 2003, followed by
the $50 note in late 2004 and the $10 in March
2006 (the $5 note is anticipated in early 2008).
These new designs include a number of security
and accessibility features that are visible and
useful without machine assistance. For visual
accessibility, the Series 1996 design, beginning
with the $50 note, introduced an enlarged, 14 mm
tall denomination numeral in the lower right

corner of the notes’ reverse side.34 The presidential
portrait also was enlarged. A machine-readable
feature to facilitate development of handheld
scanning devices for the blind and visually
impaired has been included in all new-design
banknotes issued since 1997. The Series 2004
design introduced an additional accessibility
feature, subtle background colors:
For the first time since the Series 1905 $20
Gold Certificate, the new currency featured
[sic] subtle background colors, beginning with
the new $20 note on October 9, 2003. In this
series, different colors are used for different
denominations. This will help everyone—
particularly those who are visually impaired—
to tell denominations apart. The $20 note
features subtle background colors of green,
peach and blue, as well as symbols of freedom
representing icons of Americana—in the case
of the $20 note, images of the American eagle.

It is of interest to compare the features of
these newer banknotes designs to recommendations made in the previous 1983 and 1995 studies.
In its most recent report, the NRC (NRC, 2006,
p. 10) concludes that these features “do not provide adequate differentiation for many visually
impaired individuals and provide no method of
differentiation for blind persons.” This perhaps
was not unexpected. Putting aside the enlarged
portrait, the Series 1996 design added a single
truly usable feature for the visually impaired:
the enlarged, 14 mm denomination numeral in
the lower right corner of the reverse. But, the
NRC previously noted that a numeral this size is
near the visual limit for someone with 20/400
acuity when viewed at a reading distance of 16
inches (NRC, 1995): The NRC recommended a
numeral more than twice as large. As an example,
the NRC cites studies regarding denomination
recognition of current $1 and $10 banknotes
displaying 12 mm-tall numerals on their fronts,
which standard acuity tests suggest should be
34

33

The BEP refers to additional series (for example, Series 1999,
2001, and 2003) to denote changes in the signatures on the notes;
these were not design changes, however. See BEP (2004) and
www.moneyfactory.gov/section.cfm/4/27/.

388

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

2007

U.S. banknotes typically display denomination in numerals of
between 8 mm and 12 mm tall. On $1 and $10 notes, for example,
the upper front numerals are approximately 12 mm tall, while the
numerals on the lower front and on the reverse are approximately
8 mm tall. The $100 note does not have the large high-contrast
numeral.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

adequate for many visually impaired—in fact,
users with measured acuity less than 20/60 have
difficulty recognizing such numerals at the distances of 16 to 24 inches that are common in retail
cash transactions. With respect to color, the NRC
(NRC, 1995) recommended that at least one side
of each denomination’s design feature a single
prominent color, not subtle shades of background
colors. Specifically, they wrote:
Most other countries also use different predominant colors for each denomination. The
simple detection color is faster than finding
and reading printed numbers, especially for
those with poor letter acuity. It should be
noted, however, that many people with low
vision have difficulty in discriminating subtle
shades of color and that color cues are generally
less obvious at low levels of illumination.
Consequently, any color features should use
clearly distinguishable colors. (NRC, 1995,
p. 25)

We display Series 2004 notes in Figure 11
and leave final judgment to the reader. Neither
NRC report, however, addressed the trade-off
between improved visual accessibility and concerns for counterfeit deterrence and widespread
recognition and acceptance, especially outside
the United States. We are not aware of any study
that addresses this trade-off.
Current improvements in technology portend
new features that will be of benefit to the blind
and visually impaired. The NRC’s 2007 report
(NRC, 2007), commissioned by the BEP, concludes
that advances in currency substrates will be the
only effective future deterrent to counterfeiting
as improving digital technology eases image
reproduction. But the report also contains fascinating glimpses of new substrates that might permit designers to include innovative features to
assist the visually impaired denominate currency.
“Shape memory” materials, for example, are ones
that allow an object’s shape to change when subject to stress and return to the original shape when
stress is released. Similarly, embedded piezoelectric crystals allow an object’s features to
change when an electric current is applied. Conceptually, at least, such materials could allow
embedded tactile denomination markings—say,
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

similar to those in Canadian and Japanese notes—
to “appear on demand” when a note is stretched
or when it is energized by a battery operated
device. Unlike current tactile markings, which
degrade with abrasion, such markings perhaps
could last the lifetime of the note. Technology
also suggests the possibility of notes that change
color when subject to an electric current, aiding
both counterfeit deterrence and the long-sought
goal of varying banknotes’ primary colors by
denomination: For most users, the appearance of
U.S. banknotes would be unchanged, while the
visually impaired could energize special visual
features as required. More analysis of such materials and features will be required before they
could be implemented, and an international education program might be required if the “look and
feel” of U.S. banknotes is affected by changes in
substrates. But the promise intrigues.

WHAT DO OTHER COUNTRIES
DO? SECURITY AND
ACCESSIBILITY FEATURES
Banknote designers worldwide seek to include
features both to deter counterfeiters and assist
the visually impaired. Most banknotes incorporate several security features, including holograms,
microprinting, serial numbers, and watermarks.
Most notes also include one or more visual accessibility feature, including different sizes and principal colors for different denominations, oversize
high-contrast numerals, and tactile features such
as embossed numerals or recessed/raised dots.
Among major currency issuers worldwide, only
banknotes issued by Canada and the United States
do not differ in size by denomination. In this
section, we explore the features of currency issued
by nine industrialized countries and the euro zone.
In their 1995 study, the NRC surveyed 171
currency-issuing authorities (NRC, 1995). In many
countries, security and accessibility features were
regarded as complementary. More than 120 countries varied note sizes by denomination, 24 featured oversized denomination numerals, and 167
used different primary color schemes for different
denominations. Among larger countries, only the
S E P T E M B E R / O C TO B E R

2007

389

Williams and Anderson

Table 2
Summary of Currency Design Features Intended To Assist the Visually Impaired, Major Issuing
Authorities

Issuing authority

Primary colors
that differ by
denomination*

Size that
differs by
denomination

Tactile
recognition
symbols¶

Certain
oversize
numerals#

Machine-readable
features for
banknote readers

Australia

Y

Y

Y

Y

N

Canada

Y

N

Y

Y

Y

Area†

Y

Y

Y

Y

N

Hong Kong SAR

Y

Y

Y

Y

N

Japan‡

Y

Y

Y

N

N

Singapore

Y

Y

Y

Y

N

Sweden

Y

Y

Y

Y

N

Switzerland

Y

Y

Y

Y

N

Y

Y

N

Y

N

Y/N

N

N

Y

Y

Euro Currency

United Kingdom
United States§

NOTE: *In this table, we define the primary color of the banknote as that used for the prominent portrait (if any), borders, and the
denomination numerals, or, in some cases, the dominant color of the note. Because we are not aware of any current cross-country
study comparing the color design features of banknotes, judgments are those of the authors. †For the 13 member states of the European
Union participating in the single currency. ‡Bank of Japan notes use less intense, near-pastel colors. Although each denomination uses
a different dominant color for the portrait, background, and numerals, some readers of previous versions of this article felt that the
differences are not sufficient to support our conclusion. We disagree. §Currently issued Series 2004 U.S. banknotes in denominations
of $10, $20, and $50 feature different dominant colors. The U.S. Bureau of Engraving and Printing has announced that new designs for
the $5 and $100 notes also will feature differing dominant colors. All currently issued notes except the $1 note feature a large denomination numeral on the reverse. The $1 note features a prominent “ONE” in the center of the reverse. ¶Including both embossed symbols
and patterns, and special high-relief intaglio printing (if cited by issuing authority as a visual accessibility feature). #Included if cited
by issuing authority as a visual accessibility feature.

United States and Canada did not vary the size
of their notes by denomination.
Table 1 compares and contrasts banknotes’
characteristics relative to those of an “ideal currency.” Table 2 narrows the comparison to five
visual accessibility features: colors that vary by
denomination, sizes that vary by denomination,
inclusion of tactile recognition symbols, oversize
denomination numerals, and machine readability
features for banknote readers. (The features shown
in the tables are as reported by the respective
monetary authorities in public documents. We
have not undertaken any independent scientific
investigation regarding the features of the banknotes.) In short, we find that no currency includes
all five features. Seven of the ten currencies
include four features, and three currencies include
three features.
390

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

2007

Currency images reproduced in this article
are as accurate as can be achieved with modern
high-volume printing but do not necessarily match
colors in the original banknotes. For some notes,
to deter counterfeiters, designers select colors
that cannot be accurately reproduced by office
copiers, printers, and high-volume offset presses.

Australia35 (see Figure 2)
The Reserve Bank of Australia is responsible
for printing, issuing, reissuing, and canceling
Australian notes (currency name: Australian
dollar; currency code36: AUD; currency symbol:
35

Additional information may be found at www.rba.gov.au/
CurrencyNotes/.

36

Currency codes for each country discussed correspond to the
International Organization for Standardization ISO 4217 threeletter codes. See www.iso.org/iso/en/CatalogueDetailPage.
CatalogueDetail?CSNUMBER=34749&ICS1=3&ICS2=60&ICS3=.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

$ or AU$). (Notes are printed by Note Printing
Australia Limited, a subsidiary of the Reserve
Bank.) Since 1996, all Australian banknotes have
been printed on a polymer substrate.37 The substrate is manufactured by Securency Pty Ltd., a
joint venture between the Reserve Bank and
Innovia Films. The substrate is created by coating
a clear polypropylene film with an opacifying ink.
The “press ready” substrate is supplied to Note
Printing Australia. Traditional security printing
processes are used to print the notes, including
offset and intaglio. As a final step, a clear overcoat is applied to the notes. It is possible to achieve
the same intaglio height on polymer substrate as
on paper, by depositing a greater quantity of ink.38
When “blind embossing” polymer (using an
uninked intaglio plate), a greater degree of emboss
can be achieved in the substrate.
According to the Reserve Bank of Australia,
the counterfeiting rate for Australian banknotes
is “very low,” averaging between five and six
counterfeits per million notes in circulation in
recent years. Almost all counterfeits of Australian
notes are printed on paper rather than polymer.
Yet, even polymer notes are not immune to
counterfeiting attempts; in March 2006, Interpol
seized more than AU$5 million of such notes in
Colombia, a popular location for counterfeiters.39
Such attempts appear to be rare; so far as we are
aware, no such notes were placed into circulation.
Although more costly than paper notes,
polymer notes are more durable (they last at least
four times longer), more difficult to counterfeit,
and recyclable. Australian banknotes vary in
color and length according to denomination.
The AU$5 banknote is the shortest (130 mm by
37

38

39

Since 1998, New Zealand banknotes have been produced in
Australia by the same process.
The height of the relief on polymer notes has encountered some
controversy. With paper substrate, a somewhat higher initial relief
may be obtained than on polymer, but advocates of polymer substrates have argued that the higher relief degrades quickly to a
lower relief than on polymer notes. We take no position on this,
leaving debates regarding the exotic technical details of intaglio
printing on alternative substrates to others. See, for example,
Coventry (2001), only one of many papers on this topic.
See www.interpol.int/Public/FinancialCrime/CounterfeitCurrency/
recentEvents.asp or the BIS-G10 Central Bank Counterfeit
Deterrence Group (CBCDG) joint counterfeit deterrence task force,
www.rulesforuse.org/pub/index.php?lang=en&page=pressreleases.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

65 mm), and the AU$100 banknote is the longest
(158 mm by 65 mm).
Security Features
Among the security features included in the
banknotes are
• Clear window with “blind” embossing and
a printed image
• See-through registration device
• Shadow image (similar to a watermark)
• Intaglio printing
• Background print (including screen traps)
• Microprinting
• Machine-readable features including UV
fluorescence, infrared, and magnetics/inks.
Accessibility Features
The Reserve Bank of Australia consulted with
a number of entities representing the visually
impaired in designing the polymer note series.
Among the accessibility features of the banknotes are
• Lengths that vary by denomination
• Strong color contrasts that vary by
denomination
• Bolder numerals
Raised intaglio printing of denomination
numerals, portraits, and other elements is both a
security and accessibility feature. In addition,
Blind Citizens Australia provides, free of charge,
a credit-card size device to measure note length
and distinguish between denominations. So far
as we are aware, no handheld reader is available
that uses the included machine-readable features
to assist the visually impaired.

Canada (see Figure 3)
The Bank of Canada has sole responsibility
for issuing, designing, producing, and distributing
Canadian banknotes (currency name: Canadian
dollar; currency code: CAD; currency symbol $
or C$).40 The Bank of Canada’s first banknote
series was printed on a combination of flax and
40

Additional information may be found at www.bankofcanada.ca/
en/banknotes/.

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

2007

391

Williams and Anderson

Figure 2
Australia: Security and Accessibility Features

5

Front

4
3
Security Features

1
2

1. Clear window
2. See-through registration device
3. Shadow image
4. Microprinting
Accessibility Features
5. Bolder numerals
• Length differentials by
denomination
• Strong color contrasts

1
2
Back

NOTE: See Tables 1 and 2 and the appendix for a complete listing of security and accessibility features.
SOURCE: Reserve Bank of Australia.

392

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

Figure 3
Canada: Security and Accessibility Features
Front

6
5
Security Features
1. Metallic strip
(holographic stripe)
2. Ghost image (watermark)
3. Dashes (security thread)
4. See-through number
5. Raised ink (intaglio printing)
Accessibility Features
5. Large, high-contrast numerals
6. Tactile feature
• Varying color by denomination

Back

NOTE: See Tables 1 and 2 and the appendix for a complete listing of security and accessibility features.
SOURCE: Bank of Canada.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

393

Williams and Anderson

cotton paper with the same nominal composition
as U.S. banknote paper. Since 1983, however,
Canadian banknotes have been printed on 100
percent cotton paper to conform to Quebec environmental laws on flax use.
Canadian banknotes are uniform in size but
vary in color by denomination. Notes measure
approximately 6 inches long by 2.75 inches
wide.41
The production cost of Canadian banknotes
is approximately 9 Canadian cents per note. The
estimated life of the banknotes by denomination
is as follows:
• C$5: 1 to 2 years
• C$10: 1 to 2 years
• C$20: 2 to 4 years
• C$50: 4 to 6 years
• C$100: 7 to 9 years

Through the Canadian National Institute for
the Blind, blind and visually impaired persons
can obtain (at no cost) a handheld banknote reader
that signals denomination to the user via voice,
tone, or vibration.
The tactile denomination scheme, of Canadian
invention, consists of sets of raised dots that vary
in position and number according to denomination. The dot system is not Braille which, contrary to public impressions, is not known by all
the blind. The basic “symbol” in the scheme is a
tightly spaced rectangular block of six dots—the
number and placement of such blocks indicates
the denomination. The C$5 note has one block,
for example, while the C$20 note has three such
blocks separated by a smooth surface. Canadian
banknotes also include intaglio printing of the
denomination numerals, which doubles as a
security and accessibility feature.
Security Features

According to the Bank of Canada, Canadian
counterfeit notes make up a very small percentage of notes in circulation.42 The most recently
issued banknote series—entitled the Canadian
Journey series—contains both enhanced security
features to deter counterfeiting and tactile features
to aid the visually impaired.
The Bank of Canada consulted with the
Canadian National Institute for the Blind, the
Canadian Council for the Blind, and the blind
and visually impaired community in the design
of their Canadian Journey and Birds of Canada
bank note series to help the blind and visually
impaired with recognition of bank notes by sight,
touch, or electronic devices. The Bank of Canada,
in collaboration with the blind and visually
impaired community, has encouraged the development of machines to denominate Canadian notes.

• Holographic stripe
• Watermark portrait
• Windowed color-shifting thread
• See-through number when note is held
to the light
• Raised ink (intaglio printing)
• Fine-line patterns
• Microprinting
• Fluorescence, presence of elements only
visible under UV light
• Serial number
• Unique, difficult-to-reproduce colors
Accessibility Features
• Primary colors that vary by denomination
• Tactile symbols embossed and back-coated
for durability
• Large high-contrast denomination numerals

41

With the exception of the 1935 and 1937 banknote issues, which
are 6 inches by 2.875 inches.

42

The Royal Canadian Mounted Police’s Bureau for Counterfeit and
Document Examinations estimates that in 2005, the volume of
counterfeit notes passed and seized was approximately 422,489
of 1,501,743,000 notes in circulation, or 0.00028percent of the total
notes in circulation (www.rcmp.ca/factsheets/fact_counterfeit_e.
htm). The figure was even lower in 2006: According to the Bank
of Canada, 313,207 counterfeit notes were passed and seized, from
a total of 1,567,318,000 notes in circulation, or 0.00020 percent.

394

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

2007

• Machine-readability feature, free banknote
readers

Euro Currency Area (see Figure 4)
The euro banknotes (currency name: Euro;
currency code: EUR; currency symbol: €) were
introduced as legal tender in the Eurosystem on
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

January 1, 2002.43 The ECB and the national
central banks (NCBs) of the participating countries have the right to issue euro banknotes, but
in practice only the NCBs issue banknotes.44
Banknote production is pooled between the NCBs,
and the approximate cost is €0.08 per note.45
Euro banknotes are made of 100 percent cotton. Banknotes vary in height, length, and color
according to denomination. The €5 banknote is
the smallest in size, measuring 120 mm long by
62 mm wide; the €500 banknote is the largest,
measuring 160 mm long by 82 mm wide. The
estimated life of euro banknotes by denomination
is as follows46:
• €5: 1.2 years
• €10: 1.2 years
• €20: 1.4 years
• €50: 3.3 years
• €100: 7.5 years
• €200: 10.4 years
• €500: 23.1 years
The weighted average lifetime for all denominations is two years.
The ECB has consulted with the European
Blind Union since the beginning of the euro note
design phase in 1995 to ensure notes are highly
accessible to the blind and visually impaired.
Each banknote—whether €5 or €500—had
to be readily identifiable by people with residual
or no sight. For example, the differences in height

between the lower, more commonly used denominations were quite distinct. The €100, €200, and
€500 banknotes were made the same height—to
facilitate automatic handling—but repetitive
embossed patterns were placed on the €200 and
€500 banknotes to differentiate them.
Large and bold value numerals were placed
in a standard position throughout the series, on
both sides, and sharply contrasting colors were
used. They were derived principally from the
color wheel of the Swiss painter and art teacher
Johannes Itten (1888-1967). The colors chosen
for successive denominations were taken from
opposing segments of the color wheel and thus
strongly contrast with each other. The same is
true for denominations that have digits in common, like the red €10 and green €100 banknotes
(ECB, 2007).
While euro banknotes include denominationspecific machine-readable features intended to
assist authentication and denomination by third
parties (primarily high-volume currency processors and banknote equipment manufacturers), so
far as we are aware, electronic banknote readers
have not been developed to assist visually
impaired individuals.
Counterfeiting of euro banknotes is reportedly
low. For 2003-06, the ECB reports detecting and
confiscating from circulation approximately
600,000 notes annually (with approximately 10
billion in circulation) (ECB, 2006b). Figures on
notes confiscated by law enforcement before
entering circulation are confidential.
Security Features

43

44

45

46

The Eurosystem comprises the ECB and the NCBs of the European
Union member states that have adopted the euro. Currently, 13
member states of the European Union participate in the single
currency—namely, Belgium, Germany, Greece, Spain, France,
Ireland, Italy, Luxembourg, the Netherlands, Austria, Portugal,
Slovenia, and Finland.

• Made of pure cotton
• Intaglio printing
• Tactile marks (on €200 and €500)
• Wire and multi-tone watermark

Additional information may be found at
www.ecb.int/bc/euro/banknotes/html/index.en.html.
The NCBs authorize privately and publicly owned production
facilities to print banknotes, and each facility prints one or two of
the seven denominations. Once they have printed the banknotes,
the issuing central banks exchange them. This process increases
efficiency and quality control. The ECB monitors the stocks of the
banknotes.
Data furnished by the ECB. Life span of the banknotes is calculated
by dividing, for each denomination, the average circulation of
banknotes during the previous 12 months, by the number of banknotes determined to be unfit during the previous 12 months.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

• Security thread
• See-through register when note is held to
the light
• Perforations in the hologram foil
• Hologram foil
• Optically variable ink (€50, €100, €200,
and €500)
• Iridescent stripe (€5, €10, and €200)
S E P T E M B E R / O C TO B E R

2007

395

Williams and Anderson

Figure 4
Euro Currency Area: Security and Accessibility Features

Front

6
5
3

Security Features
1. Holographic image and perforations

4

2. Color-changing ink
3. Security thread

1

4. Watermark
5. Partial images printed on the front
and back align to form the value
numeral (see-through number)
6. Raised ink (intaglio printing)
Accessibility Features

5

• Different sizes: the higher the
banknote, the bigger the banknote
• Striking, clearly contrasting colors
• Large numerals
• Intaglio printing

2
Back

NOTE: See Tables 1 and 2 and the appendix for a complete listing of security and accessibility features.
SOURCE: European Central Bank.

396

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

• Mini and microprinting
• Fluorescence, presence of elements only
visible under UV light
Accessibility Features
• Sizes that vary by denomination
• Primary colors that vary by denomination:
bold, striking colors/clear, contrasting
colors among denominations
• Oversize denomination numerals
• High-relief intaglio printing
• Tactile marks (€200and €500 only)

Hong Kong (see Figure 5)
The Hong Kong Monetary Authority has
authorized (delegated) the issuance of Hong Kong
banknotes (currency name: Hong Kong dollar;
currency code: HKD; currency symbol: $ or HK$)
to three commercial banks—The Hongkong and
Shanghai Banking Corporation Ltd., the Standard
Chartered Bank, and the Bank of China (Hong
Kong) Limited— under specific terms and conditions stipulated by the Hong Kong government.
The notes issued by the three banks differ somewhat in design although all banknotes are printed
in Hong Kong by Hong Kong Note Printing
Limited.47 The notes illustrated here are those
issued by HongKong and Shanghai Banking
Corporation Ltd. (HSBC).
Under normal usage, a banknote is estimated
to last for an average of about two years. Production costs are confidential.
In 2003, the Hong Kong Monetary Authority
reported that:

Although more recently, there has been a
problem of counterfeits of the old series (2000
and 2002) Hongkong and Shanghai Banking
Corporation $1,000 banknotes.49
Hong Kong banknotes are made of 100 percent
cotton fiber,50 which does not fluoresce under
UV light. They vary in length and color according
to denomination. The HK$10 banknote is the
smallest (133 mm by 66 mm), and the HK$1,000
banknote is the largest (163 mm by 81.5 mm).
Enhanced intaglio printing serves both as a security and accessibility feature.
In the design phase for the currency issued
in 2003 and 2004, the Hong Kong Monetary
Authority consulted with major advocacy organizations to include accessibility features for the
blind and visually impaired. In addition, the three
note-issuing banks and the Hong Kong Monetary
Authority produce a template to aid the visually
impaired in measuring a banknote’s size.
Security Features
• Fluorescent machine-readable barcode
• Iridescent images
• Denomination numeral in optical variable
ink
• Holographic windowed thread
• Concealed image/denomination
• Intaglio printing
• Invisible fluorescent fibers
• Watermarks
• Security thread
• See-through feature
• Serial numbers

“Forgery of banknotes is not a major problem
in Hong Kong, thanks to the quality of our
existing banknotes and the good work of the
police in tracking down counterfeiters. However, constant changes in technology are
increasing the vulnerability to counterfeiting
and it is thus necessary to keep more than one
step ahead of those who might try to exploit
any gaps in the security of our banknotes…”48
47

Additional information may be found at www.info.gov.hk/hkma/
new_hk_banknotes/eng/index.htm.

48

See www.info.gov.hk/hkma/eng/viewpt/20030904e.htm.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Accessibility Features
• Sizes that vary by denomination
• Colors that vary by denomination
• A bold, large denomination numeral on
the front of all banknotes at a common
49

See www.info.gov.hk/hkma/eng/press/2007/20070320e3.htm.

50

The HK$10 polymer banknote is scheduled to be put into circulation gradually starting July 9, 2007. See www.info.gov.hk/hkma/
eng/press/2007/20070615e6.htm and www.info.gov.hk/hkma/
eng/polymer/leaflet_polymer_eng.pdf.

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

2007

397

Williams and Anderson

Figure 5
Hong Kong: Security and Accessibility Features

3

Front

1
Security Features
1. Fluorescent machine-readable
barcode
2. See-through feature
3. Denomination numeral in optical
variable ink

2

4. Holographic windowed thread
Accessibility Features
• Different sizes
• Distinctive colors
• Large, bold denomination
numerals
• Enhanced intaglio printing

4

Back

NOTE: See Tables 1 and 2 and the appendix for a complete listing of security and accessibility features.
SOURCE: The Hong Kong and Shanghai Banking Corporation Limited, with permission of the Hong Kong Monetary Authority.

398

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

location, with deep engraving to give a
pronounced tactile feel

studies to improve the accessibility features for
the visually impaired.

• A bold denomination numeral in negative
outline located at a common location at
the back of all banknotes

Security Features
• Hologram (¥10,000 note)
• Watermark bar pattern

• Enhanced intaglio printing to give stronger
embossing effect

• Latent image

• Banknote size measurement template

• Pearl ink
• Ultra-fine-line printing

Japan (see Figure 6)

• Intaglio printing

The Bank of Japan is the sole issuer of Japanese
banknotes (currency name: yen; currency code:
JPY; currency symbol: ¥ ). The banknotes are
manufactured by the National Printing Bureau,
an incorporated administrative agency.51
In response to the increase of Japanese counterfeit banknotes in recent years, the Bank of
Japan issued a new series of banknotes in 2004
with improved security features. There are four
denominations: ¥1,000, ¥2,000, ¥5,000, and
¥10,000. These redesigned Japanese yen notes vary
in width and color according to denomination.
The ¥1,000 banknote is the shortest note, measuring 76 mm wide by 150 mm long, and the ¥10,000
banknote is the longest, measuring 76 mm wide
by 160 mm long. Bank of Japan notes contain
machine-readable features intended to assist
mechanical note processing. So far as we are
aware, no consumer-level devices to assist the
blind and visually impaired have been developed.
The average durability of the Bank of Japan
notes is one to two years for the ¥1,000 and ¥5,000
notes and four to five years for the ¥10,000 notes.52
In fiscal year 2005, the total manufacturing
cost of 4.08 billion banknotes amounted to ¥61.7
billion, approximately ¥15.1 per banknote.
The current series of Bank of Japan notes
contain “tactile marks” created by high-relief
intaglio printing to assist the visually impaired.
High-relief numerals, created by intaglio printing,
both deter counterfeiting and assist the visually
impaired. The Bank of Japan occasionally makes

• Microprinting

51

Additional information may be found at www.boj.or.jp/en/type/
release/zuiji/kako03/bnnew3.htm.

52

Source: Bank of Japan.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

• Luminescent ink
Accessibility Features
• Tactile marks (intaglio printing)
• Sizes that vary by denomination
• Colors that vary by denomination

Singapore (see Figure 7)
The Monetary Authority of Singapore (MAS)
is the sole issuer of banknotes (currency name:
dollar; currency code: SGD; currency symbol: $
or S$) in Singapore.53 In 1990, the MAS issued
their first polymer note of S$50 denomination.
Polymer notes are more durable and cost-effective.
Both paper and polymer banknotes are currently
in circulation. Their “Portrait Series” notes were
first issued in 1999. This series has seven denominations. The banknotes vary in length, width, and
color according to denomination. The smallest
note is the S$2 note, measuring 126 mm in height
by 63 mm in width; the largest note is the
S$10,000 note and measures 180 mm in length
by 90 mm in height. Singapore banknotes include
machine readable features intended to assist currency processing and authentication.
High-relief intaglio printing of several characteristics, including the denomination numeral,
serves as an accessibility and anti-counterfeiting
feature. During the design phase of the Portraitseries notes, MAS consulted with the Singapore
Association of the Visually Handicapped on
their needs for the new series of banknotes. The
53

Additional information may be found at www.mas.gov.sg/
masmcm/bin/pt1Singapore_Circulation_Notes.htm.

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

2007

399

Williams and Anderson

Figure 6
Japan: Security and Accessibility Features

1

5

5

2

Front

7

4

4
Security Features

6

1. Hologram
2. Watermark bar
pattern

1

8

5

5

3

3. Latent image

8

4. Pearl ink
5. Microprinting
6. Luminescent ink

1

7. Intaglio printing

3

Accessibility Features
8. Intaglio-printed
recognition symbol

2

• Different sizes

5

• Different colors

4

5

5

5

Back

NOTE: See Tables 1 and 2 and the appendix for a complete listing of security and accessibility features.
SOURCE: Bank of Japan.

400

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

Figure 7
Singapore: Security and Accessibility Features

10

11

9

3

15

1

2
8

8

7

3

4

6
2

12

5

Paper Note

Front

7

Back

1

6

6

15

2
3
10

14

2

13

6

13

Polymer Note
Security Features
1.
2.
3.
4.
5.

Engraved portrait
Asymmetrical serial number
Anti-color copying line structures
Lift twin
Intaglio printing

15. Braille codes

6.
7.
8.
9.
10.

Perfect registration
Highlight watermark
Watermark
Kinegram
Microprinting

Accessibility Features
• Different sizes

11.
12.
13.
14.

Lithographic print
Security thread
Invisible feature
Fluorescent ink

• Different colors

NOTE: See Tables 1 and 2 and the appendix for a complete listing of security and accessibility features.
SOURCE: Monetary Authority of Singapore.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

401

Williams and Anderson

Association recommended the printing of Braille
codes and varying note size by denomination. The
Braille codes are printed in heavy, high-relief
intaglio on the top right of the front of each note.
According to the Monetary Authority of
Singapore, the cost and duration of their banknotes
by denomination are as follows:
• S$2 polymer
S$0.10
72-96 months (issued in January 2006,
still monitoring the performance)
• S$5 paper
18-24 months

S$0.10

• S$10 paper
24-36 months

S$0.09

• S$10 polymer S$0.12
72-96 months (issued in May 2004, still
monitoring the performance)
• S$50 paper
24-36 months

S$0.08

• S$100 paper
48-60 months

S$0.07

• S$1,000 paper S$0.09
48-60 months
• S$10,000 paper S$0.55
48-60 months
Security Features
• Microprinting

Accessibility Features
• Braille codes
• Sizes that vary by denomination
• Colors that vary by denomination
• Intaglio printing

Sweden (see Figure 8)
The Sveriges Riksbank is the sole issuer of
Swedish banknotes (currency name: krona; currency code: SEK; currency symbol: kr).54 Swedish
banknotes are 100 percent cotton. Swedish banknotes are issued in five denominations (20, 50,
100, 500, and 1,000 kronor) and vary in length,
width, and color according to denomination. The
20 kronor banknote is the smallest in size, measuring 67 mm in width and 120 mm in length. The
1,000 kronor banknote is the largest, measuring
82 mm in width and 160 mm in length. Banknote
lifetime varies by denomination, with 20 kronor
banknotes being fit for circulation for about one
year and 1,000 kronor for about five years. The
cost of production is about 7.5 U.S. cents per
banknote. Swedish banknotes contain machinereadable features but, so far as we are aware, no
handheld devices are available to assist the visually impaired. Intaglio printing of the denomination and portrait serve as both security and
accessibility features.

• Kinegram
Security Features

• Lithographic print

• Watermark

• Engraved portrait

• Security thread

• Braille codes

• See-through picture

• Asymmetrical serial numbers (vertical and
horizontal)

• Intaglio printing

• Anti-color-copying line structures

• Foil strip/hologram

• Lift twin
• Intaglio printing

• Fluorescent picture (for banknotes with a
foil strip)

• Highlight watermark

• Microlettering

• Watermark

• Shimmering mother-of-pearl ink

• Perfect registration

• Motion (a moving image in the striped
bank image; included in the 1,000 krona
note with a foil strip only)

• Security thread
• Invisible feature: non-reflective under UV
light
• See-through windows on polymer notes
402

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

2007

54

Additional information may be found at www.riksbank.com/
templates/SectionStart.aspx?id=10890.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

Figure 8
Sweden: Security and Accessibility Features

Front

6
1
2
3
4

5
6

6
Security Features
1. Foil strip with hologram
2. Watermark
3. Security thread
4. See-through picture
5. Raw cotton banknote, gives
the note a special texture
Accessibility Features
6. Intaglio printing
• Different sizes
• Different colors
Back

NOTE: See Tables 1 and 2 and the appendix for a complete listing of security and accessibility features.
SOURCE: Riksbank.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

403

Williams and Anderson

Figure 9
Switzerland: Security and Accessibility Features

1
2
Security Features
1. Transparent image
2. Fine lines that change
colors

9

3. Microtext
4. Security thread
5. Serial number

11

6. Tilt effect
7. Metallic digits

8

8. Optically variable print
9. Intaglio printing

4

Accessibility Features
10. Recognition symbol

6

11. Perforated number

5

7

•

Different sizes

•

Different colors

3

10
Front

Back

NOTE: See Tables 1 and 2 and the appendix for a complete listing of security and accessibility features.
SOURCE: Swiss National Bank.

404

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

Accessibility Features

• Optically variable ink

• Sizes that vary by denomination

• UV digits

• Colors that vary by denomination

• Metallic digits

• Intaglio printing

• Tilt effect
Accessibility Features

Switzerland (see Figure 9)
The Swiss National Bank’s (SNB) network of
cash distribution services is the issuing and
redeeming authority of Swiss banknotes (currency
name: Franc; currency code: CHF; currency symbol: CHF, Fr).55 The network includes the SNB’s
three bank offices, 16 agencies, and 234 domestic
correspondents. Swiss franc notes are printed on
paper made of linters and short cotton fibers.
Swiss franc banknotes vary in length and color
according to denomination. There are six denominations (10, 20, 50, 100, 200, and 1000 francs).
All notes measure 74 mm in width. The length
varies by denomination, in 11 mm increments,
with the 10 franc banknote being the shortest
(126 mm) and the 1000 franc being the largest
(181 mm). Some of the security elements in Swiss
banknotes are machine readable.
The cost of producing Swiss franc banknotes
is approximately 30 centimes per new note. The
average duration of the banknotes is as follows:
• 10 CHF: 2 years
• 20 CHF: 1.5 years
• 50 CHF: 1.5 years
• 100 CHF: 3.5 years
• 200 CHF: 3 years
• 1000 CHF: 10 years
Perforated numerals serve as both an accessibility and security feature. A tactile symbol varying by denomination is embossed at the lower end
of each note as an additional accessibility feature.

• Sizes that vary by denomination
• Colors that vary by denomination
• Perforated number
• Intaglio printing (digits)

United Kingdom (see Figure 10)
The Bank of England is the sole issuer of
banknotes (currency name: pound; currency
code: GBP; currency symbol: £) in the United
Kingdom.56 Bank of England banknotes are produced by De La Rue Currency and made from
cotton fiber and linen rag. The banknotes vary
by color, length, and width according to denomination, with the £5 note measuring 135 mm by
70 mm and the £50 note measuring 156 mm by
85 mm.
The Bank of England estimates an average
life of banknotes of about 1 year for the £5 note
to five or more years for the £50 note.
During the banknote design phase, advice
from the Royal National Institute of the Blind, a
U.K. charity, is sought and their recommendations
are implemented. For the new £20 series F note,
for example, the Royal National Institute of the
Blind recommended that large-high contrast
numerals be included and the size and color differential between denominations be retained;
both were accepted by the Bank of England. In
addition, a recognition symbol unique to each
denomination is included on the front of the
banknote to aid the visually impaired.
Security Features

Security Features
• Made of cotton and linen paper
• Transparent register

• Metallic thread

• Watermark digits

• Watermark

• Intaglio printing (digits)

• Print quality

• Perforated number
56
55

Additional information may be found at www.snb.ch/en/iabout/cash.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Additional information may be found at www.bankofengland.co.uk/
banknotes/current/index.htm.

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

2007

405

Williams and Anderson

Figure 10
United Kingdom: Security and Accessibility Features

Front

8
1
5

Security Features
1. Intaglio printing

4

2. Metallic thread
3. Watermark
4. Print quality

6

5. Holographic strip
6. Microlettering
7. See-through register
Accessibility Features
8. Large denomination numerals

2

7

• Different sizes
• Different colors

3

Back

NOTE: See Tables 1 and 2 and the appendix for a complete listing of security and accessibility features.
SOURCE: Bank of England.

406

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

• Hologram

• $1: 21 months

• UV light

• $5: 16 months

• Microlettering

• $10: 18 months

• Foil (on £50 note)

• $20: 24 months

Accessibility Features

• $50: 55 months

• Sizes that vary by denomination

• $100: 89 months

• Colors that vary by denomination
• Large denomination numerals
• Densely colored recognition symbols to
indicate denomination:
• £5: turquoise circle
• £10: orange diamond
• £20: purple square
• £50: red triangle

United States (see Figure 11)
The United States government is the sole
issuer of U.S. banknotes (currency name: dollar;
currency code: USD; currency symbol: $).57 The
BEP, a part of the U.S. Treasury Department, produces U.S. banknotes, which are placed into circulation by the Federal Reserve Banks. Banknotes
are made of paper composed of 25 percent linen
and 75 percent cotton, with red and blue synthetic
fibers distributed evenly throughout the paper.
U.S. banknotes do not vary by size or primary
color across denominations; they measure 2.61
inches wide by 6.14 inches long and 0.0043 inches
thick (or 66.3 mm wide × 155.96 mm long and
0.11 mm thick). There are seven denominations
of U.S. banknotes ($1, $2, $5, $10, $20, $50, and
$100). The $100 note has been the largest denomination issued since 1969. The $2 banknote circulates in small numbers.
The estimated cost of producing U.S. banknotes was 5.6 cents per banknote during fiscal
year 200658. The estimated average life of a U.S.
banknote varies by denomination:
57

58

The BEP prints all banknotes, based on production orders sent by
the Federal Reserve Banks, and ships the notes to the Federal
Reserve: at this point, they are just paper. The notes are placed
into circulation by the Federal Reserve Banks, which means the
Fed ships them from their vaults to a non-Fed location. Additional
information may be found at www.moneyfactory.gov/.
www.federalreserve.gov/generalinfo/foia/2007newcurrency.htm.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

The U.S. government introduced the “New
Currency Design” U.S. banknotes in 1996. This
was the first major redesign of U.S. banknotes
since 1928. Sophisticated security features were
added to deter counterfeiting domestically as
well as abroad. Security features include colorshifting ink, watermarks, and security threads. In
2003, The “New Color of Money” banknotes were
introduced. These banknotes feature different
subtle background colors for different denominations to assist the visually impaired. A larger
numeral in the lower right corner on the back of
the notes was included as an additional accessibility feature. All new Federal Reserve notes
issued since 1997 include machine-readable features (such as infrared markings) that might be
used by handheld currency readers for the visually
impaired. To date, few such readers have been
introduced and some users have reported difficulty using them with newer design banknotes
(see footnote 25).
Security Features
• Watermark
• Color-shifting inks
• Fine-line printing patterns (intaglio printing
on both the face and back)
• Enlarged off-center portraits
• Low-vision feature
• Machine-readable feature
• Security thread
• Microprinting
Accessibility Features
• Larger denomination numeral on reverse
• Large portraits
• Subtle color patterns on face and back
(Series 2004)
S E P T E M B E R / O C TO B E R

2007

407

Williams and Anderson

DETAIL OF SELECT SECURITY FEATURES
FOR THE $50 U.S. BANKNOTE

NOTE: The above insert is a partial reproduction of an image from the web site of the U.S. Bureau of Engraving and Printing.
The site provides numerous images, information, and tutorials on the characteristics of U.S banknotes:
www.moneyfactory.gov/newmoney/files/Sec_Features_50.pdf.

408

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

Figure 11
United States: Security and Accessibility Features
Front

1
5
5

4

Security Features

5

2

1. Enlarged, off-center portrait

6

2. Subtle, over-printed colors

7

3

3. Color-shifting ink
4. Fine-line printing patterns
5. Microprinting
6. Watermark
7. Security thread
Accessibility Features
8. Large denomination numeral on
reverse

5

• Enlarged portrait
• Colors that vary by denomination

8
Back

NOTE: See Tables 1 and 2 and the appendix for a complete listing of security and accessibility features.
SOURCE: U.S. Bureau of Engraving and Printing.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

409

Williams and Anderson

SUMMARY AND CONCLUSIONS
Currency is widely used in the payment systems of all countries. In each system, there are a
large number of participants—households, retail
merchants, vending machine firms, financial
institutions, high-volume currency processors,
and the central bank or currency-issuing authority.
The objectives of these stakeholders need not
necessarily coincide, and some difficulties in
settling on an acceptable currency design for the
nation is to be anticipated.
In this analysis, we have focused on currency
design features that enhance counterfeit deterrence and improve accessibility for the visually
impaired. All currencies issued worldwide contain some of these design features but, interestingly, different countries have made different
choices. Among banknotes worldwide, only U.S.
notes are widely held outside their country of
issue. These foreign holdings benefit the United
States when interpreted as a type of non-interestbearing U.S. government debt, and simultaneously
benefit foreign holders by providing a stable store
of value and medium of exchange in often unstable countries. As a result, the effectiveness of U.S.
banknotes’ counterfeit deterrence features is a
matter of concern worldwide.
All banknotes contain features to deter counterfeiting. Some features, including paper or
polymer substrate and intaglio printing, assist the
public to readily authenticate notes by touch;
indeed, it is widely reported that counterfeit notes
are most often detected by the public and bank
tellers because “they don’t feel right.” For authentication, these tactile features are valuable both
to normally sighted and visually impaired users.
Other note features, including microprinting,
magnetic inks, color-varying fluorescent threads,
and infrared components, are valuable for machine
authentication across a wide range of devices,
from handheld devices and vending machines to
high-speed currency sorters. Recently published
analyses argue that continuing advances in printing technology will require banknote designers
to work aggressively to remain ahead of sophisticated counterfeiters.
410

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

2007

All banknotes also contain features to assist
the visually impaired. Published studies have
concluded that the single feature of greatest value
for the visually impaired in denominating notes
is note sizes that vary by denomination. More
than 180 countries in the world issue banknotes.
Only two major currency-issuing authorities issue
notes that do not vary in size by denomination—
Canada and the United States. Some nation’s
have selected designs in which bright primary
colors vary by denomination—Canadian and euro
notes, for example—while in others, such as the
Series-2004 U.S. banknotes and Bank of Japan
notes, subtle background colors vary by denomination. Most current note designs include larger
denomination numerals than previous designs—
Canadian and euro notes feature perhaps the most
prominent numerals. Series-2004 Federal Reserve
notes feature an enlarged denomination numeral
on the reverse, albeit smaller than the Canadian
or euro note numerals.
The goal of our analysis has been to increase
public awareness of the variety, and the trade-offs,
among currency design features worldwide. We
will have been successful if each reader appreciates a bit more the banknotes they use every day.

REFERENCES
Andrews, Edmund L. “U.S. Currency Discriminates
Against the Blind, Judge Rules,” The New York
Times, November 29, 2006.
Bank for International Settlements. “General Guidance
for National Payment System Development.”
Committee on Payment and Settlement Systems,
January 2006.
Board of Governors of the Federal Reserve System.
Ninety-Second Annual Report, 2005. Washington,
DC, 2006a.
Board of Governors of the Federal Reserve System.
2007 New Currency Budget. Washington, DC,
December 2006b; www.federalreserve.gov/
generalinfo/foia/2007newcurrency.htm.
Bureau of Engraving and Printing. A Study of
Mechanisms for the Denomination of U.S. Currency

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

by the Blind or Visually Impaired. Report to
Congressman Edward R. Roybal, Chairman, U.S.
House of Representatives Select Committee on
Aging, 1983; fraser.stlouisfed.org/historicaldocs/
tresbep/download/30353/study_of_mechanisms.pdf.
Bureau of Engraving and Printing. “U.S. Unveils New
$50 Note with Background Colors.” April 28, 2004;
www.moneyfactory.gov/newmoney/
main.cfm/media/releases04262004.
Coventry, Les. “Polymer Notes and the Meaning of
Life.” Presented at the international currency
conference, on behalf of the Reserve Bank of
Australia, Barcelona, Spain, April 2001;
www.rba.gov.au/CurrencyNotes/ConferencePapers/
cu_coventry_0401.html and
www.currencyconference.com.
de Heij, Hans A.M. “Public Feedback for Better
Banknote Design.” Proceedings on behalf of De
Nederlandsche Bank NV at IS&T/SPIE’s 18th Annual
Symposium on Electronic Imaging; Conference No.
6075, Optical Security and Counterfeit Deterrence
Techniques VI, San Jose, CA, January 18, 2006; version 2 dated February 1, 2006; www.dnb.nl/
dnb/home/file/Public%20feed%20back%20for%20
better%20banknote%20design_tcm46-145677.pdf.
de Heij, Hans A.M.; DiNunzio, Lisa A. and Strube,
Olivier. “Comparence EUR–USD Banknotes: Public’s
Appreciation and Knowledge.” Presented on
behalf of De Nederlandsche Bank NV to the
Banknote 2003 conference, Washington, DC,
February 2-5, 2003; www.dnb.nl/dnb/home/file/
Comparence%20EURxUSD%20banknotes%
20Banknote%202003_tcm46-145670.pdf.
European Central Bank. Annual Report, 2005. April
2006a; www.ecb.int/pub/annual/html/index.en.html.
European Central Bank. Biannual Information on Euro
Banknote Counterfeiting. Press release, July 13,
2006b; accessed July 2, 2007 at www.ecb.int/press/
pr/date/2006/html/pr060713.en.html.
European Central Bank. “How the Euro Became Our
Money: A Short History of the Euro Banknotes and
Coins.” 2007; www.ecb.int/pub/pdf/other/
euro_became_our_moneyen.pdf.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Europol. An Overview of Forgery of Money. January
2006a; www.europol.europa.eu/publications/
SeriousCrimeOverviews/2005/sc6.pdf.
Europol. EU Organized Crime Threat Assessment
2006, 2006b; www.europol.europa.eu/publications/
OCTA/OCTA2006.pdf.
Lambert, Michael J. and Staton, Kristin D.
“Opportunities and Challenges of the U.S. Dollar
as an Increasingly Global Currency: A Federal
Reserve Perspective.” Federal Reserve Bulletin,
September 2001; www.federalreserve.gov/pubs/
bulletin/2001/0901lead.pdf.
Lang, Brian. “Effectiveness of Public Security and
Features on Bank Notes.” Speech to Interpol, Tenth
International Conference on Currency Counterfeiting,
Reserve Bank of New Zealand, April 2002;
www.rbnz.govt.nz/speeches/0116878.html.
Mihm, Stephen. “No Ordinary Counterfeit.” New
York Times Magazine, July 23, 2006.
National Eye Institute. “More Americans Facing
Blindness Than Ever Before. Report Released on
One of the Most-Feared Disabilities.” March 20,
2002; www.nei.nih.gov/news/pressreleases/
032002.asp.
National Research Council. Counterfeit Deterrent
Features for the Next-Generation Currency Design.
Committee on Next-Generation Currency Design
and Commission on Engineering and Technical
Systems. Washington, DC: National Academy Press
of the National Academy of Sciences, 1993.
National Research Council. Currency Features for
Visually Impaired People. Committee on Currency
Features Usable by the Visually Impaired.
Washington, DC: National Academy Press of the
National Academy of Sciences, 1995.
National Research Council. Is It Real? Identification
and Assessment of the Counterfeiting Threat for
U.S. Banknotes. Committee on Technologies to
Deter Currency Counterfeiting. Washington, DC:
National Academy Press of the National Academy
of Sciences, 2006.

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

2007

411

Williams and Anderson

National Research Council. A Path to the Next
Generation of U.S. Banknotes: Keeping Them Real.
Committee on Technologies to Deter Currency
Counterfeiting. Washington, DC: National Academy
Press of the National Academy of Sciences, 2007.

United States Treasury Department. U.S. Department
of the Treasury FY 2008 Budget in Brief.
Washington, DC: U.S. Government Printing Office,
2007; www.treas.gov/offices/management/budget/
budgetinbrief/fy2008/bib.pdf.

United States Treasury Department. The Use and
Counterfeiting of United States Currency Abroad.
Washington, DC: U.S. Government Printing Office,
January 2000.

Van Renesse, Rudolf L. Optical Document Security.
Third Edition. Boston: Artech House Publishers,
2005.

United States Treasury Department. The Use and
Counterfeiting of United States Currency Abroad,
Part 2. Washington, DC: U.S. Government Printing
Office, March 2003.
United States Treasury Department. Treasury Bulletin.
Washington, DC: U.S. Government Printing Office,
March 2006a.
United States Treasury Department. The Use and
Counterfeiting of United States Currency Abroad,
Part 3. Washington, DC: U.S. Government Printing
Office, September 2006b.

Vitale, Susan; Cotch, Mary Frances and Sperduto,
Robert P. “Prevalence of Visual Impairment in the
United States.” Journal of the American Medical
Association, May 10, 2006, 295(18), pp. 2158-63.
Volcker, Paul A. Statement before the Subcommittee
on Domestic Monetary Policy of the Committee on
Banking, Finance and Urban Affairs, U.S. House of
Representatives, January 29, 1986;
fraser.stlouisfed.org/historicaldocs/pav/download/
28190/Volcker_19860129.pdf.

APPENDIX
Web Sites for Data on Currency Features, Counterfeit Activity, and
Numbers of Notes in Circulation
Australia
Currency features:
Currency Features:www.rba.gov.au/CurrencyNotes/SecurityFeaturesAndCounterfeitDetection/
security_features_on_australias_notes.htmlwww.rba.gov.au/CurrencyNotes/vision_impaired.html
Counterfeit volume:
www.rba.gov.au/PublicationsAndResearch/RBAAnnualReports/2005/Html/bus_serv.html
www.rba.gov.au/PublicationsAndResearch/RBAAnnualReports/2006/Html/bus_serv.html
Notes in circulation:
www.rba.gov.au/CurrencyNotes/Statistics/notes_on_issue.html
Canada
Currency features:
www.bankofcanada.ca/en/banknotes/counterfeit/security_features.htmlwww.bankofcanada.ca/
en/banknotes/accessibility.html
412

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Williams and Anderson

Counterfeit volume
www.rcmp.ca/factsheets/fact_counterfeit_e.htm
www.rcmp.ca/scams/counter_e.htm
Notes in circulation:
www.bankofcanada.ca/en/annual/2006/annualreport_2006.pdf
Euro Currency Area
Currency features:
www.ecb.int/bc/euro/banknotes/security/html/index.en.htmlwww.ecb.int/bc/euro/banknotes/
visually/html/index.en.html
Counterfeit volume:
www.ecb.int/press/pr/date/2007/html/pr070112.en.html
Notes in circulation:
www.ecb.int/bc/faqbc/figures/shared/number_of_banknotes_circulation_pdf_web.pdf?
Hong Kong
Currency features:
www.info.gov.hk/hkma/new_hk_banknotes/eng/index.htmwww.info.gov.hk/hkma/eng/viewpt/
20030904e.htm
Counterfeit volume:
Not available
Some discussion on counterfeiting:
www.info.gov.hk/hkma/eng/press/2007/20070320e3.htm
Notes in circulation:
www.info.gov.hk/hkma/eng/statistics/msb/new_msb_tables_b.htm#money
www.info.gov.hk/hkma/ar2005/english/economic/economic06.htm
Japan
Currency features:
www.boj.or.jp/en/type/release/zuiji/kako03/bnnew3.htm
Counterfeit volume:
www.boj.or.jp/en/type/release/teiki/ar/data/ar0605.pdf
Notes in circulation:
www.boj.or.jp/en/type/stat/dlong/fin_stat/money/cdab0010.csv
Singapore
Currency features:
www.mas.gov.sg/currency/currency_info/The_Singapore_Currency.html
www.mas.gov.sg/resource/currency/currency_info/Brochure%20on%20$2%20Polymer%20Note.pdf
Counterfeit volume:
Not available
Some discussion on counterfeiting:
www.mas.gov.sg/news_room/letters_to_editors/2007/No_reason_for_retailers_not_to_accept_
100dollars.html
Notes in circulation:
www.mas.gov.sg/currency/currency_stat/Currency_in_Circulation_Notes.html
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

413

Williams and Anderson

Sweden
Currency features:
www.riksbank.com/templates/SectionStart.aspx?id=10890
www.riksbank.com/upload/Dokument_riksbank/Kat_publicerat/Broschyrer/banknotes_coins_
060925.pdf
www.riksbank.com/templates/Page.aspx?id=23941
Counterfeit volume:
www.riksbank.com/upload/Dokument_riksbank/Kat_publicerat/Rapporter/2007/
arsred_popvers_06_eng.pdf
Notes in circulation:
www.riksbank.com/upload/Dokument_riksbank/Kat_publicerat/Rapporter/2007/
arsred_popvers_06_eng.pdf
Switzerland
Currency features:
www.snb.ch/en/iabout/cash/current/design/id/cash_current_design_principle
www.snb.ch/en/iabout/cash/current/id/cash_security
Counterfeit volume:
www.snb.ch/ext/link?url=%2fen%2fmmr%2freference%2fannrep_2004_accrep%2fsource
www.snb.ch/ext/link?url=%2fen%2fmmr%2freference%2fannrep_2005_accrep%2fsource
www.snb.ch/ext/link?url=%2fen%2fmmr%2freference%2fannrep_2006_komplett%2fsource
Notes in circulation:
www.snb.ch/ext/link?url=%2fen%2fmmr%2freference%2fannrep_2006_rechenschaft%2fsource
www.snb.ch/ext/link?url=%2fen%2fmmr%2freference%2fannrep_2006_komplett%2fsource
United Kingdom
Currency features:
www.bankofengland.co.uk/banknotes/kyb_lo_res.pdf
www.bankofengland.co.uk/banknotes/security/index.htm
Counterfeit volume:
www.bankofengland.co.uk/publications/annualreport/2005report.pdf
www.bankofengland.co.uk/publications/annualreport/2006report.pdf
March 2007 issue of Currency News: Sterling and the Bank of England—A Study in Contradictions
Notes in circulation:
www.bankofengland.co.uk/banknotes/noteissue.pdf
United States
Currency features:
www.moneyfactory.gov/section.cfm/4www.moneyfactory.gov/newmoney/
Counterfeit volume:
www.treas.gov/offices/domestic-finance/acd/
Notes in circulation:
www.moneyfactory.gov/document.cfm/18/106

414

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Experiments in Financial Liberalization:
The Mexican Banking Sector
Rubén Hernández-Murillo
Since the liberalization of its trade in the mid-1980s, Mexico has pursued an aggressive globalization strategy, which today makes it the country with the most free trade agreements in the world.
This liberalization strategy has also included the banking sector, particularly since 1997, when all
restrictions to the entry of foreign banks were removed. The history of the banking sector in Mexico
includes episodes of nationalization in 1982, privatization in 1992, and near-complete failure in
1995. Since then, however, the Mexican government has undertaken a series of bold reforms that
have contributed to the modernization of its financial system. This paper documents the evolution
of Mexico’s banking sector starting from its nationalization in 1982 and culminating with the
increased entry of foreign banks in recent years that has driven the recovery of bank credit to the
private sector. (JEL G18, G21, G28, P11, D23)
Federal Reserve Bank of St. Louis Review, September/October 2007, 89(5), pp. 415-32.

I

n the mid-1980s Mexico started to liberalize its trade; and, since the signing of the
North American Free Trade Agreement
(NAFTA) in early 1994, Mexico has followed an aggressive globalization strategy, placing about 90 percent of its trade flows under free
trade agreements with over 40 countries. These
polices have made Mexico the country with
the most free trade agreements in the world.1
Mexico’s liberalization strategy has also included
its financial sector and, in particular, the banking
industry.
Mexico’s experience with financial liberalization provides an interesting case study for at least
two reasons. First, economic theory suggests that
financial liberalization bolsters economic growth.
Mexico’s path toward financial liberalization has
been an arduous one and includes several failed
attempts, which, until recently, prevented the
1

At the time of this writing, Mexico was negotiating a free trade
agreement with Mercosur to join this bloc as an associate member.

development of its banking sector and limited the
growth of financial credit to the private sector,
which is necessary for economic development.
Second, given the nature of some of the
experiments that have been followed, Mexico’s
experience can also provide lessons about the
effectiveness of aggressive openness reforms targeted at improving competition in the banking
industry and at increasing credit to the private
sector—namely, the elimination of all restrictions
on foreign ownership of banking assets in 1997,
which allowed foreign banks to dominate the sector. One of the most recent developments in this
regard is the approval in 2006 of Wal-Mart’s bid
to create a commercial bank in Mexico.
In this paper I describe the evolution of
Mexico’s banking sector by reviewing (roughly
in chronological order) the reforms that have
been implemented since 1982, when Mexico’s
financial system was nationalized. Continuing
problems with Mexico’s institutional and legal
framework persisted even after the banking sector

Rubén Hernández-Murillo is a senior economist at the Federal Reserve Bank of St. Louis. The author thanks Gabriel Cuadra at Banco de
Mexico for help in obtaining the data used for the figures. Marcela Williams and Christopher Martinek provided research assistance.

© 2007, The Federal Reserve Bank of St. Louis. Articles may be reprinted, reproduced, published, distributed, displayed, and transmitted in
their entirety if copyright notice, author name(s), and full citation are included. Abstracts, synopses, and other derivative works may be made
only with prior written permission of the Federal Reserve Bank of St. Louis.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

415

Hernández-Murillo

Figure 1
Credit by Commercial Banks
Share of GDP
0.450

Total
Private Sector

0.400

0.350

0.300

0.250

0.200

0.150

0.100

0.050
1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

Quarter (end of period)

SOURCE: Banco de México.

was re-privatized in 1992 and generated distorted
incentives among lenders who engaged in risky
lending practices. These practices ultimately led
to a banking crisis by 1995, forcing the Mexican
government to implement a massive rescue program to prevent generalized failure in the sector.
Since 1997, however, the Mexican government has undertaken a series of bold reforms: for
example, extensive reforms to banks’ accounting
standards, elimination of all barriers to the entry
of foreign banks, reforms to personal and business
bankruptcy laws, improvements in credit rating
technology and regulatory changes to promote
the use of credit bureaus, reforms to capitalization
and risk management standards that conform with
the recommendations of the Basel II accords, and,
more recently, reforms that promote the securitization of mortgage loans. By the end of 2006,
banks in Mexico were owned almost entirely by
foreign financial institutions, with subsidiaries of
416

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

2007

foreign banks representing more than 80 percent
of total banking assets in the system.
These reforms have started to produce positive results and promise even more favorable
changes in the years ahead. The most positive
result thus far is that the credit crunch that followed the banking crisis of 1995 appears to
have subsided; and bank credit to the private
sector has been growing rapidly since 2004. (See
Figure 1.)

NATIONALIZATION OF THE
BANKING SYSTEM, 1982
At the end of the administration of President
López Portillo, in 1982, Mexico experienced its
most severe crisis since 1932, with real gross
domestic product (GDP) declining by 4.7 percent
in the fourth quarter of 1982. (See Figure 2.) The
crisis was triggered by adverse shocks to oil prices
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Hernández-Murillo

Figure 2
Real GDP
Year-Over-Year Growth Rate (%)
10
8
6
4
2
0
–2
–4
–6
–8
–10

19
82
19
83
19
84
19
85
19
86
19
87
19
88
19
89
19
90
19
91
19
92
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
20
06

–12

Quarters

SOURCE: Banco de México.

and world interest rates, but was ultimately caused
by a disorderly expansionary fiscal policy largely
supported by borrowing from international capital
markets and by borrowing from the central bank
(i.e., simply printing money to finance the deficit),
which the government had followed since the
1970s. The government devalued the currency
and defaulted on foreign debt payments, which
caused Mexico to lose access to international
credit markets.
In September 1982, the government also
announced a decree to amend the constitution
and nationalize 58 of the 60 banks in the sector.
The exceptions were Citibank, which has had a
presence in Mexico since 1929, and Banco Obrero,
which was owned by a labor union.2
During the new administration of President
Miguel de la Madrid Hurtado, the banking system
was consolidated into a smaller number of banks,
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

but remained heavily regulated otherwise. The
number of commercial banks was reduced to 29
in 1983, to 20 in 1985, and finally to 18 in 1988.
In the early 1980s, the banking sector was the
dominant financial intermediary and, along with
the stock market, it constituted the core of the
financial system. However, its role was diminished
throughout the nationalization episode as the
government restricted universal banking activities,
prohibiting banks from forming integrated financial groups to offer other financial services beyond
banking; consequently, other types of financial
intermediaries began to take on more important
roles.3 Private investors turned to alternative
sources of financing, such as commercial paper,
2

See Bubel and Skelton (2002).

3

With the re-privatization of the banking sector in 1992, banks
would again operate under a universal banking structure.

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

2007

417

Hernández-Murillo

further reducing the role of banks as financial
intermediaries.
From 1982 to 1984, for example, brokerage
firms essentially controlled the sale and purchase
of financial instruments, which included commercial paper and bonds.4 Although short-term
Mexican Treasury bills had been introduced since
1978, the interest rate on these instruments was
set exogenously. This policy limited the extent
of their use to finance the public deficit. In 1982,
however, a new auction mechanism was developed for their exchange, allowing the market to
determine the rate of return.5
In 1984, the new administration began to sell
off bank operations that did not take deposits and
make loans. The goal was to separate different
financial intermediation roles among independent
types of operations to allow private investors
(often former bank owners) to form and operate
stock brokerage firms, insurance and re-insurance
firms, and currency exchange firms.6
After being excluded from international
credit markets, from 1982 to 1988, the Mexican
government turned to domestic credit to finance
its fiscal deficits, imposing heavy reserve requirements on the newly nationalized banks. As a
result, bank lending to the private sector declined
dramatically.7 The government also imposed
many other constraints on the banking sector,
which induced lending inefficiencies, including
interest rate ceilings on bank deposits and loans,
as well as mandated lending quotas on highpriority economic sectors.8
Commercial banks could freely allocate only
up to 25 percent of their credit, for which they
could charge market lending rates; but they were
required to allocate all other loans to the federal
government or to targeted sectors of the economy.
Furthermore, until 1989, the government also
controlled banks’ operational procedures, such
as deciding on the location and number of bank
4

See Trigueros (1995).

branches that could be created and the number
of employees that were hired and approving the
banks’ annual budgets for income and expenses.

FINANCIAL LIBERALIZATION
REFORMS, 1988-89
Starting in 1988, a new series of reforms were
initiated to make the financial system more competitive. Among these reforms was the elimination in April 1989 of controls on interest rates and
the sectoral quotas imposed by the government
on commercial lending. The reserve requirements
on private deposits were eliminated in 1991.
These reforms culminated with the re-privatization of the banking sector in 1991 and 1992 under
the administration of President Carlos Salinas de
Gortari.
By this time, the Mexican money market had
become more liquid and the government increased
the issuance of short-term Treasury bills so that
it no longer relied on commercial bank financing.9
Mexico had also regained entry to international
capital markets by 1989 after the implementation
of the Brady Plan and enactment of a stabilization
plan (begun in 1987).10 Bank credit to the private
sector began to increase dramatically after 1989.11
Development banks, that is, state-controlled
banks charged with providing credit to private
and state-owned enterprises in targeted sectors
in the economy, also changed significantly. The
share of loans allocated to state-owned firms
decreased and loans were increasingly allocated
to private enterprises. At the same time, however,
total development bank credit to nonfinancial
firms in the private sector decreased as the regulations on commercial banking were removed.12
During this period, the regulatory framework
of the financial system was updated and new laws
9

See Gruben and McComb (1997).

10

The Brady Plan, named after U.S. Treasury Secretary Nicholas
Brady, was the mechanism used to restructure bonds issued by
many Latin American countries that defaulted on their debt in the
1980s. Mexico was the first country to repay its Brady debt, in 2003.

5

See Katz (1990).

6

See Gruben and McComb (1997).

7

See Gruben and McComb (1997).

11

See Gelos and Werner (2002).

See Gelos and Werner (2002).

12

See Gonzalez-Anaya and Marrufo (2001).

8

418

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Hernández-Murillo

were drafted to regulate commercial banks and
insurance and bonding companies, as well as the
stock market and investment banks. In 1990, the
Salinas administration created a new deposit
insurance authority, FOBAPROA (Banking Fund
for the Protection of Savings).
In July 1990, the government passed laws
allowing for the formation of integrated financial
groups that could consolidate different types of
financial intermediation under a scheme of universal banking. The law allowed for the integration of
banks with leasing, factoring, currency exchange,
mutual fund management, and asset-based warehousing firms; and, separately, it allowed for the
integration of brokerage firms with leasing, factoring, currency exchange, mutual fund management,
and asset-based warehousing firms.13 The final
group allowed was holding companies.14 Initially,
however, restrictions remained for the integration
of commercial banks with brokerage firms and
insurance and bonding firms.15

THE PRIVATIZATION OF THE
BANKING SECTOR, 1990-92
The privatization of commercial banks was
part of a broad program of privatization of staterun enterprises under a general stabilization program that included increased fiscal discipline
and cuts in government spending, as well as
inflation-reduction measures initiated by the de
la Madrid administration. The constitutional
amendment allowing for the re-privatization of
the banking sector was announced in May 1990.
In August of that year, the finance ministry
announced the principles that would inspire the
privatization process. Two of the goals of the program were to increase competition and efficiency
in the financial system and to improve bank
capitalization.
13

Asset-based warehousing firms, or almacenes generales de
depósito, stored goods in exchange for certificates of deposit,
which could be sold to other credit institutions. See Katz (1990).

14

See Unal and Navarro (1999).

15

See Katz (1990).

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

The government took great efforts to guarantee transparency of the process; they indicated
that the sale price of the banks would be consistent with valuations based on objective criteria.
But the results of these valuations, however, were
not made public, and the overvaluation of some
banks became obvious after the purchase was
completed.
Some studies underscore several problems
with the financing schemes that buyers adopted
to purchase these banks from the government,
which in some cases included borrowing from the
same banks they were buying.16 These studies
seem to agree that the privatization auctions were
designed to maximize the purchase price of the
banks.
The auctions were conducted from June 1991
through July 1992. Interested parties were allowed
to acquire the controlling shares of only one bank,
and the privatization was open only to Mexican
nationals. There were six rounds of bidding.
Packages of three or four banks were sold in each
round. The sequencing of the sale was designed
to increase competition in subsequent stages of
the bidding process. The largest banks were sold
in the first rounds. Interestingly, the government
provided no minimum bid or reserve price
information.17 At the end of the process, banks
were sold at an average price-to–book value ratio
of 3.04, producing about $12.4 billion for the
government.18
Some studies suggest that buyers paid those
prices anticipating a concentrated market structure
with oligopolistic profits. In fact, of the existing
18 banks, the 4 largest banks controlled 70 percent of total bank assets, and Mexican banks did
not face competition from the entry of foreign
banks, as they were not allowed to participate in
the privatization auctions.19
The lack of a competitive market structure in
the newly privatized banks was not the only prob16

See Unal and Navarro (1999) and Haber (2005b).

17

See Avalos and Hernández Trillo (2006).

18

The calculation of the average is weighted by assets. See Murillo
(2002) and Unal and Navarro (1999).

19

See Haber (2005b,c).

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

2007

419

Hernández-Murillo

lem. More fundamental problems plagued the
institutional framework under which the banks
were to operate. From the outset of the new privatization era, bankers faced incentives that would
encourage inefficient, risky lending.
Haber (2005b,c), for example, argues that at
the time of the privatization, Mexico had weak
property rights institutions to assess the credit
worthiness of borrowers and to enforce the contract rights of bankers. Furthermore, he also argues
that out of the privatization process emerged a
set of institutions that reduced the incentives of
bank directors, bank depositors, and bank regulators to enforce the prudent behavior of the newly
privatized banks. Some of these institutions
would not be reformed until after the 1995 collapse. This view is shared by Unal and Navarro
(1999), who argue that, despite the recent reforms,
the new banks operated under an outdated regulatory environment and that the supervisory
agencies were often unable to implement newly
adopted regulations or to enforce existing rules.
Efficient monitoring both internally and by
the government was also lacking. No regulatory
body required the banking sector to adhere to
the generally accepted accounting principles in
international markets. Accounting standards
were very lax and, in particular, did not require
banks to report the entire value of past-due loans
as nonperforming, but only the past-due interest
payments; banks were allowed to roll over the
principal of those loans. Banks were not required
to provide consolidated financial reports until
1995, even though at this time they were operating
under a universal banking structure.20 This lack
of regulation made it difficult to establish limits
on lending within financial groups. In addition,
the Mexican central bank implicitly guaranteed
unlimited deposit insurance through FOBAPROA.
There were no credit bureaus, and legal institutions in Mexico did not provide an adequate
enforcement of lending contracts.
In contrast with the expectations of an oligopolistic market structure, banks soon were competing for market share.21 Credit was growing very

rapidly, and nonperforming loans soon became a
problem.

20
21

THE BANKING BAILOUT
Risky Lending Practices
The December 1994 devaluation and macroeconomic crisis that ensued was not the origin of
the banking crisis; it was merely the trigger.22
Induced by the lack of an appropriate legal and
regulatory environment, banks had been engaged
in risky lending practices almost immediately
after the privatization process was concluded
and past-due loans had been rising dramatically
prior to the devaluation.
Several factors—including low inflation,
improvements in the fiscal surplus of the government, an exchange rate under a controlled rate of
depreciation, and low real interest rates—contributed to an expansion of both bank credit and
financial intermediation in general. The size of
the financial sector, given by M4 as a share of
GDP, reached 47 percent at its peak in 1994.23
(See Figure 3.)
From December 1988 to November 1994,
credit from local commercial banks to the private
sector rose in real terms by 277 percent, or 25 percent per year.24 Nonperforming loans, however,
grew even faster. Between December 1991 and
December 1993 alone, gross past-due loans more
than tripled in absolute terms, while the share of
reported nonperforming loans to total loans rose
from 4.13 to 7.26 percent.25 This situation per22

After a failed attempt of a controlled devaluation on December 20,
1994, a run against the peso led to the collapse of the semi-fixed
exchange rate regime and the peso was allowed to float on
December 22, 1994. For a detailed account of the events and an
analysis of the factors contributing to the vulnerability of Mexico’s
financial sector, see Calvo and Mendoza (1996) and Gil-Díaz (1998),
and also Gruben and McComb (1997).

23

M4 is a broad monetary aggregate defined as M3 + deposits in
branches of domestic banks abroad, where M3 = M2 + domestic
financial assets held by non-residents, M2 = M1 + domestic financial assets held by residents, and M1 = currency outside of banks,
domestic and foreign currency checking accounts in resident banks,
domestic and foreign currency current account deposits in resident
banks, and sight deposits in savings and loan associations.

See Gonzalez and Marrufo (2001).

24

See Gil-Díaz (1998).

See Gruben and McComb (1997).

25

See Gruben and McComb (2003).

420

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Hernández-Murillo

Figure 3
Size of the Financial Sector: M4
Share of GDP
0.60

0.55

0.50

0.45

0.40

0.35

0.30
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Quarter (end of period)

SOURCE: Banco de México.

sisted into the bailout episode that followed.
Haber (2005b) estimates that, as of December 1991,
the ratio of nonperforming loans to total loans
(including principal rollovers and the value of
loans transferred to FOBAPROA) was 13.5 percent; it increased to 17.1 percent by December
1994, to 36.3 percent by December 1995, and to
52.6 percent by December of 1996.
An example of risky lending behavior was
related lending, a topic studied by La Porta, López
de Silanes, and Zamarripa (2003). Related lending
refers to the practice of lending to separate firms
in which bank officials have interests either
because they own these other firms or their associates or family members do. In their study, the
authors tracked a sample of loans outstanding at
Mexican banks from the end of 1995 through 1999.
They found that about 20 percent of loans were
to related parties. They also found that these
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

parties were more likely to borrow at lower rates,
were less likely to post collateral, and were more
likely to default than unrelated parties. The
authors examined the regulations set forth after
the 1990 privatization and noted that very few
rules addressed conflicts of interest arising from
lending to related parties. The authors identified
other key factors, beyond regulation of conflict of
interests, that provided incentives to engage in
risky lending policies: the universal guarantees
provided by FOBAPROA (which guaranteed all
deposits up to 100 percent, regardless of the
creditworthiness of the bank and the total amount
of deposits) and the minimal capitalization
requirements imposed on the banks (which were
unrelated to the riskiness of the bank’s loan portfolio). So, because of the presence of deposit insurance, the lower a bank’s capital, the greater its
incentive was to take on additional risk because
S E P T E M B E R / O C TO B E R

2007

421

Hernández-Murillo

Figure 4
Inflation
Annualized Change (%)
490.0
440.0
390.0
340.0
290.0
240.0
190.0
140.0
90.0
40.0

19
82
19
83
19
84
19
85
19
86
19
87
19
88
19
89
19
90
19
91
19
92
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
20
06
20
07

–10.0

SOURCE: Banco de México.

less of their own wealth was at stake.26 La Porta,
López de Silanes, and Zamarripa (2003) argued
that banks engaged in excessive related lending
because of moral hazard: Legally, bank assets
were considered separate from the related firm’s
assets and regulatory authorities were insuring
these loans, too.27

The Rescue Program
The devaluation of December 1994 generated capital outflows and high inflation. Real
GDP declined by 6.2 percent in 1995, while the
annualized inflation rate reached 46.9 percent
in December 1995. (See Figures 2 and 4, respectively.) The central bank imposed restrictive credit
26

See Furlong and Keeley (1989) and Keeley (1990).

27

For additional literature on the moral hazard problems generated
by deposit insurance institutions, see Demirgüç-Kunt and
Detragiache (2002), Cull, Senbet, and Sorge (2005), and also Martin
(2006), as well as references therein.

422

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

2007

and monetary policies in February 1995, and
interest rates skyrocketed. Many borrowers were
unable to repay their loans. The rising level of
past-due loans had put the banking system at the
brink of total collapse.28
Three regulatory authorities intervened in
the design and implementation of the rescue
package that started in 1995 and continued
through December 1998. The first was Mexico’s
central bank, Banco de México. The second authority was Mexico’s finance ministry, Secretaría de
Hacienda y Crédito Público, which is the main
financial authority in Mexico ultimately responsible for regulating and supervising the entire
financial system. These tasks, however, are delegated to the third regulatory authority, the national
banking and securities commission, Comisión
Nacional Bancaria y de Valores, which regulates
28

See Gruben and McComb (1997) and McQuerry (1999).

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Hernández-Murillo

and supervises all financial intermediaries in
Mexico.29
The Mexican government adopted several
measures to prevent the collapse of the banking
system, including efforts to improve banks’
immediate liquidity requirements following the
December 1994 devaluation, measures to improve
capitalization, various debtor relief efforts, and
ultimately the transfer of nonperforming loans
into public debt, approved by the Mexican
Congress in December 1998. The funding of these
programs was channeled through the nation’s
bank deposits insurance institution, whose liabilities amounted to about $60 billion in February
1998 near the end of the rescue program and represented almost five times the amount received
by the government when banks were privatized.30
I now review some of the specific programs
included in the rescue package that followed the
1995 crisis.
Liquidity Window. After the December 1994
devaluation, several banks were unable to roll
over their dollar-denominated liabilities. During
1995, the central bank offered short-term dollardenominated credit through FOBAPROA so that
banks could meet immediate dollar liabilities.
The terms of these loans were designed so banks
could soon find alternative funding sources. The
program served its purpose, and all banks that
received these loans repayed them.
Initial Capitalization Program. In February
1995, the government created a special short-term
recapitalization program that would allow banks
to raise capital by selling 5-year convertible
bonds to FOBAPROA. The goal was to help
banks increase their capital-to-assets ratio above
8 percent while they tried to raise additional
capital on their own.31 The program included
reforms that removed restrictions on the ways
banks could increase capital. Banks were charged
higher interbank interest rates and were prohib29

Prior to 1995, the banking and securities commissions were two
separate agencies.

30

See Hernández Trillo and López Escarpulli (2001).

31

This was the minimum capitalization ratio recommended by the
Basel I accords.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

ited from issuing other subordinated debt until
they exited the program. The program was not
entirely successful, as many banks avoided participation, motivated in part because the market
seemed to interpret participation as a sign of
weakness and imminent regulatory intervention
by the authorities.32 Many banks tried to raise
capital on their own and failed to attain the
capitalization requirements.
Loans for Bonds Swaps. The core program
in the rescue package had the goal of preventing
generalized failure in the banking system. With
this program, FOBAPROA initially acquired only
a portion of past-due loans from commercial
banks and acquired the rights to any payments
that could be recovered by the banks.33 In
exchange, banks had to purchase FOBAPROAissued special 10-year non-negotiable bonds
backed by Mexico’s central bank. Banks were
also required to raise new capital at the rate of
one peso for each two pesos of loans transferred
to FOBAPROA. Additionally, FOBAPROA bonds
were indexed in a new unit of account (referred
to as UDI or Unidad de Inversión) created in
1995, which was indexed to the inflation rate to
guarantee the real value of funds. When the
maturities of these bonds were reached, banks
would be allowed to either roll over the debt or
sell it. Mortgage loans were also included in the
rescue package.
Participating banks agreed to surrender their
institutions to banking authorities if they were
unable to convert their debt with FOBAPROA
into equity capital.34 Some banks were unable to
raise additional capital, and FOBAPROA undertook further capitalization efforts to help these
banks. These efforts allowed banks to continue
to operate. When FOBAPROA determined that a
bank was no longer viable, the bank would be
liquidated and its assets would be sold. Banks
acquiring these assets would also take over the
bank’s liabilities, which were backed by matching
loans from FOBAPROA.
32

See Mackey (1999).

33

Later, as past-due loans continued to rise, more loans were also
included in the rescue package.

34

See McQuerry (1999).

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

2007

423

Hernández-Murillo

The national banking commission had the
right to formally intervene in cases of irregular
practices, which ranged from failure to meet
reserve and capitalization requirements to illicit
operations and fraud: The authorities would take
control of the bank’s management and suspend
stockholder privileges while the investigations
were ongoing. Meanwhile, FOBAPROA continued
to support these banks in their capitalization
efforts. By June 1998, the authorities had intervened in the operations of 12 banks. Two of these,
Banca Cremi and Banco Unión, were under the
authorities’ control prior to the peso devaluation.
Additionally, several banks, including Banca
Serfín, Mexico’s third largest bank, underwent
de facto intervention and were subjected to a more
strict supervisory process, which included many
of the controls used in a formal intervention.35
NAFTA and Foreign Banks. As part of the
rescue program, early in 1995, the government
also allowed foreign investors to purchase a
troubled bank, if that bank accounted for no more
than 6 percent of total Mexican bank capital; but
these foreign entities were not allowed to start
new banks yet. This reform legalized the purchase
of all but the three largest banks.36 The new law
also raised the maximum amount of banking
capital that could be controlled by foreign banks
to 25 percent, compared with the previous limit
of 9 percent initially allowed by NAFTA in 1994.
Debt Relief Programs. Throughout the 199598 banking bailout episode, there were several
debt relief programs designed to provide support
to small borrowers (individually or in specific
industry sectors) and help them repay their bank
loans. Some programs were financed by the
government and others involved negotiations
among the banks and their borrowers. All these
programs were supervised by the national banking commission.
These programs included features such as
reduced interest rates, payment discounts, and
debt restructuring into the newly created inflationindexed accounting unit with a fixed real interest
rate. Debtors were responsible for repaying the
35

See Mackey (1999).

36

See Gruben and McComb (1997).

424

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

real interest rate, while the government covered
the difference between the nominal and real
interest rates.
Mackey (1999) found that these debt relief
programs succeeded in helping a large number
of small debtors repay their loans. The overall
impact on these programs on total past-due loans
was most probably minimal. However, Mackey
(1999) argues that in designing these programs,
the authorities were more interested in preventing generalized runs against the banking system
caused by lack of depositors’ trust.

POST-CRISIS FINANCIAL REFORMS
Since 1995 the national banking commission
has sought the advice of U.S. government agencies,
such as the Federal Reserve and the Office of the
Comptroller of the Currency, as well as international organizations, such as the World Bank and
the International Monetary Fund, to improve its
regulatory practices (Mackey, 1999). The Mexican
government has also striven since 1997 to align
these practices with international standards,
particularly with regard to banking accounting
standards, capitalization requirements, and credit
risk qualifications; it has also enacted reforms
since then to improve bankers’ incentives, including changes to FOBAPROA and reforms to bankruptcy and mortgage laws, as well as the promotion
of private credit bureaus.

Bank Accounting Standards and
Supervision Practices
New bank accounting standards in Mexico
were approved by the banking regulatory authority in December of 1995 and were adopted in
January 1997.37 The new criteria imposed greater
disclosure on banks and made their balance
sheets more directly comparable with those in
other countries, particularly with regard to the
disclosure of nonperforming loans. The new rules
also required consolidated balance sheets and
income statements for financial groups.
37

2007

See Mackey (1999), McQuerry and Espinosa (1998), and Del
Ángel, Haber, and Musacchio (2006).

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Hernández-Murillo

As related earlier, one of the problems that
led to the banking crisis of 1995 was the lack of
strict disclosure rules regarding nonperforming
loans. Banks calculated nonperforming loans
with a due payments criteria; that is, only loan
payments 90-days past due were recorded as
nonperforming, as opposed to classifying as nonperforming the value of the entire loan itself. All
the while, the outstanding balance could still
accrue interest. In addition, if loans were restructured or renegotiated, the loan was reclassified
as a new loan. As a result of these practices, a
bank’s income statement and its capital adequacy
ratios did not reflect accurate information about
the bank’s financial health.38
Under the new laws, the value reported as
past due was now the total unpaid balance of the
loan, and loans were listed as nonperforming
after a set number of payments went unpaid. The
new criteria also required more strict provisions
for loan losses, and interest could no longer be
accumulated.
The new laws also set stricter standards to
deal with related lending. In particular, a bank
was required to inform the banking regulatory
authority of all the details of a related loan, after
the loan was approved by a majority of the bank’s
directors. Limits were also set on the total amount
of loans made to related parties.39
In 1999, the banking authorities also established new rules for the assessment and measurement of risks; these rules were strengthened in
2003 to conform with recommendations on capital
adequacy and contingency reserve requirements
from the Basel II accords.

to other international deposits insurance institutions, such as the Federal Deposit Insurance
Corporation in the United States. This new institution is designed to offer more protection to
small depositors and allows larger depositors to
face more risk. In particular, whereas FOBAPROA
offered implicitly unlimited guarantees on
deposits, IPAB offers limited guarantees explicitly.
IPAB also has the authority to formally intervene
if it detects irregularities in a member bank,
although the authority of the national banking
commission supersedes that of the IPAB.
In 2005, regulations established an upper
bound on the total amount of insured deposits
of 400,000 UDIs (about $100,000 at the thenprevailing exchange rate) per individual or business by bank.40 In general, the reason to limit
deposit-insurance coverage is to encourage depositors with large balances to monitor bank behavior
and thereby establish a better link between
deposit interest rates and balances and a bank’s
risk taking.

Bankruptcy Laws

In December 1998, Mexico’s deposit insurance
institution, FOBAPROA, was replaced by a new
entity, the Institute for the Protection of Bank
Savings, or IPAB. This institution is now responsible for insuring bank deposits and managing
bank support programs. It differs from FOBAPROA
in several aspects and is more similar in design

Before the re-privatization, Mexican laws on
bankruptcy and debt moratory dated back to 1943.
A reform in 1988 created special bankruptcy
courts to deal with business bankruptcies, which
fell within the purview of federal legislation;
personal bankruptcies, in contrast, were legislated
at the state level and dealt with in state civil
courts. In spite of these reforms, bankruptcy cases
in Mexico continued to be notoriously lengthy
and complex, as the laws generally provided little
incentives for the repayment of debts. As a consequence, lenders often faced difficulties recovering assets from insolvent borrowers. In addition,
there were not enough such bankruptcy courts.
Banks often settled negotiations with debtors
outside of the courts, with great advantages to
the latter.41
On April 25, 2000, a new law was approved
by the lower house of the Mexican Congress that
would replace existing bankruptcy laws for per-

38

See Gil-Díaz (1998).

40

See Haber (2005a).

See Mackey (1999).

41

See Mackey (1999).

Reforms to FOBAPROA

39

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

425

Hernández-Murillo

Figure 5
Bank Credit to Private Sector
Share of GDP
0.120
0.115
0.110
0.105
0.100
0.095
0.090
0.085
0.080
0.075
0.070
0.065
0.060
0.055
0.050
0.045
0.040
0.035
0.030
0.025
0.020
0.015
0.010
0.005
0.000
1994 1995

Share of GDP
0.045
Private Industry (left scale)
Private Housing (left scale)
Private Consumption (right scale)

0.040
0.035
0.030
0.025
0.020
0.015
0.010
0.005
0.000

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

Quarter (end of period)

SOURCE: Banco de México.

sons and businesses. It seems, however, that this
new law still falls short in terms of providing
banks with adequate repayment guarantees.42

Mortgage Markets
After the 1995 crisis, mortgage lending by
commercial banks plummeted along with total
private lending, and banks all but exited the housing lending market. (See Figure 5.) New financial
associations with limited functions (known in
Mexico as SOFOLES) had been created in 1993
to foster competition in the financial system; after
the 1995 crisis, these non-bank financial intermediaries became important players in the lowincome mortgage market.
These non-bank banks borrowed funds from
government development banks as well as from
42

See Hernández Trillo and López Escarpulli (2001) and Avalos and
Hernández (2006).

426

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

2007

private commercial banks; in turn they gave loans
to consumers to purchase homes and autos but
were initially prohibited from accepting deposits
or investing in securities or derivatives.43
In recent years, the role of non-bank banks in
providing financing services to the private sector
has increased remarkably. From 2000 to 2006 the
share of total loans to the private sector provided
by non-bank banks increased from 6.3 percent to
9.1 percent. In real terms, total loans by these
intermediaries grew by about 130 percent.44 By
December 2006, there were 56 non-bank banks.
The largest share of total loans granted by these
institutions has been for private housing, reaching
about 42 percent by December 2006. The nextlargest categories are loans to private industry
43

See Haber and Musacchio (2005).

44

Data through September 2006. Source: Secretaría de Hacienda y
Crédito Público.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Hernández-Murillo

(37 percent) and for private consumption (18
percent).45
Mortgage markets in Mexico are currently
undergoing a very remarkable transformation.
From 2001 to 2006, the average compound annual
growth rate of housing loans was 27.3 percent;
in 2006, these loans amounted to about $79.6
billion.46 Bank lending for private housing began
to increase in 2004. (See Figure 5.)
From 1995 to 2002, two government institutions provided most of the funding to non-bank
banks; since about 2003, however, mortgage
origination by non-bank banks and commercial
banks has been funded largely by other sources.47
Non-bank banks, investment banks, and the
Federal Mortgage Association had been pushing
forward the securitization of mortgage-backed
bonds to develop a secondary market, which
started in 2003, when the first mortgage-backed
security was issued for $178 million.48
In 2006, the first commercial bank entered
this market. Also in 2006, a new unified property
registry began operations in Mexico. Its objective
has been to concentrate records for all the existing
housing supply as well as new starts that can be
purchased with funding by private and public
mortgages lenders. By maintaining updated information, which can be used by developers, lenders,
and consumers, the registry is expected to help
improve efficiency and boost the development
of housing and mortgage markets.
In 2007, Mexico’s Federal Mortgage Association, in association with the Netherlands Development Finance Company, created a new company
called HiTo to provide a bridge between mortgage
lending and the bond market by standardizing
mortgage-backed securities into large pools, with
the goal of increasing liquidity and efficiency in
the market.49
45

Source: Comisión Nacional Bancaria y de Valores.

46

See www.shf.gob.mx/files/pdf/Estrategia%202007-2013.pdf.

47

The first government mortgage institution is a trust for lowerincome housing whose liabilities are funded by Mexico’s central
bank, the World Bank, and other sources (see Pickering, 2000). The
second institution is the Federal Mortgage Association created in
2001 to promote the development of a secondary mortgage market.

48
49

Credit Bureaus
A credit bureau operated by the central bank
has existed in Mexico since 1964, but it was only
for business loans and was rarely used.50 From
1993 to 1998, Mexican regulatory authorities laid
the framework for the existence of private credit
bureaus, which was designed to improve the way
individuals’ credit history information was collected.51 Additional laws to regulate the credit
bureaus were passed in 2002 and 2004.
The goal of credit bureaus is to improve credit
screening and assessment of repayment capabilities of individuals and firms. Starting in 1998,
the banking commission has required banks to
establish reserves for 100 percent of those loans
in which borrowers are found to have poor or no
credit history.52
Since 1995, three firms have entered the
market; but today only one, Buró de Crédito,
remains, and it now tracks the credit histories of
both consumers and businesses.

Increased Participation of Foreign
Banks
Some of the most remarkable developments
that followed the 1995 crisis were the reforms
that allowed for increased foreign ownership of
commercial banks. Today, foreign participation
has essentially rebuilt the sector, improving capitalization and the quality of bank assets and contributing to the accelerated decline of bad loans.
Since 2004, foreign participation has helped to
increase bank credit in the economy. (See Figures
1 and 5.)
In 1994, NAFTA allowed foreign-owned banks
to operate in Mexico, although there remained
severe restrictions on the market share that foreign banks were allowed to control. NAFTA also
restricted the equity share controlled by foreign
nationals on Mexican banks. By 1994 there were
35 banks in the system, and only 2 banks, which
50

See Gil-Díaz (1998).

See Skelton (2006).

51

See Negrin (2001).

See The Economist (2007).

52

See Mackey (1999).

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

427

Hernández-Murillo

Table 1
Commercial Bank Assets as of December 2006
Bank name
All banks total
BBVA Bancomer

Assets (million pesos)

Share of total

2,564,279.8

100.0

600,836.3

23.4

Foreign control

Yes

Banamex

538,881.1

21.0

Yes

Santander

402,075.2

15.7

Yes

HSBC

284,045.4

11.1

Yes

Mercantil del Norte

209,031.9

8.2

No

Scotiabank Inverlat

127,150.0

5.0

Yes

80,800.3

3.2

No

Inbursa
I.N.G. Bank

57,980.6

2.3

Yes

Del Bajío

46,131.4

1.8

No

Banco Azteca

44,088.5

1.7

No

J.P. Morgan

31,911.9

1.2

Yes

Bank of America

16,609.3

0.7

Yes

IXE

15,226.0

0.6

No

Interacciones

12,881.2

0.5

No

Afirme

11,438.6

0.5

No

American Express

10,707.5

0.4

Yes

Invex

10,345.7

0.4

No

Banregio

9,720.7

0.4

No

G.E. Capital

9,598.3

0.4

Yes

Mifel

7,581.5

0.3

No

BBVA Bancomer Servicios

5,982.1

0.2

Yes

Deutsche Bank

5,880.9

0.2

Yes

Ve por más

4,890.2

0.2

No

Bansí

4,264.5

0.2

No

Credit Suisse

3,614.5

0.1

Yes

Compartamos

3,283.8

0.1

No

A.B.N. Amro Bank

3,269.9

0.1

Yes

Barclays Bank

2,728.5

0.1

Yes

Tokio-Mitsubishi UFJ

2,402.4

0.1

Yes

Monex

569.7

0.0

No

Autofín

352.0

0.0

No

SOURCE: Assets are from Boletín Estadístico de Banca Múltiple, Comisión Nacional Bancaria y de Valores, December 2006,
www.cnbv.gob.mx.

428

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Hernández-Murillo

provided retail lending, were controlled by foreign
ownership.53
By December 1996, only 7 percent of total
bank assets were controlled by foreign banks, but
this share increased to 11 percent by December
1997 after all the restrictions were removed. By
December 1999, 20 percent of bank assets were
controlled by foreign banks; by December 2004,
83 percent of bank assets in Mexico were controlled by foreign banks.54 As of December 2006,
foreign banks still controlled about 82 percent of
total bank assets. (See Table 1.)
Haber and Musacchio (2005) argue that the
entry of foreign banks has increased competitive
pressures, which have improved efficiency in the
sector, even among those banks not controlled by
foreign banks. They find that, while foreign entry
between 1997 and 2004 improved the profitability
of the sector, foreign banks were not more profitable, on average, than domestic banks, although
foreign banks may have been initially better at
screening borrowers. Schulz (2004) has also found
that increased entry of foreign banks has helped
improve various measures of productivity in the
sector.
The entry of foreign banks in the sector has
been instrumental to the recovery of bank credit,
which had plunged after the 1994 devaluation.
Total credit by commercial banks declined from
about 40 percent of GDP in the fourth quarter of
1994 to about 13 percent in the first quarter of
2004. Similarly, bank credit to the private sector
declined from about 37 percent of GDP in the
fourth quarter of 1994 to about 9 percent in the
first quarter of 2004. (See Figure 1.)
In contrast to total bank credit, credit for private consumption started recovering soon after
the bank rescue program was completed in 1999,
reflecting in part an increase in the use of credit
cards. Other types of bank credit to the private
sector—in particular, housing loans—would not
begin to recover until 2004. Bank credit to the
private industry, however, still remains stagnant
at about 2 percent of GDP. (See Figure 5.)
53
54

The recovery in private bank credit, driven by
entry of foreign banks, has produced significant
welfare gains (in the form of lower net interest
margins) among Mexican consumers.55 The entry
of foreign banks has also generated positive
effects on banks’ productivity and capitalization
derived from the increase in foreign capital, and
the improvement in bank asset quality has helped
accelerate the reduction of bad loans in the banking system.56
FOBAPROA/IPAB bonds had begun to mature
by the end of 2005 and are now being rolled over
into negotiable bonds, which will provide banks
with additional liquidity to lend to the private
sector. The cumulative effects of Mexico’s reforms
in the financial system, including the widespread
presence of foreign banks in the sector, suggest that
the recent reversal in bank credit will continue.57

WAL-MART’S BANK IN MEXICO
In November 2006, Wal-Mart’s subsidiary in
Mexico received approval to open a bank and
became the second retail chain in Mexico to operate a bank. The first, Grupo Elektra, has been
operating a full-service bank, Banco Azteca, since
2002. In 2006, other retailers also applied for and
received licenses to operate banks. The stated
goal of both Wal-Mart’s and Elektra’s banks is to
provide banking services to Mexico’s low-income
individuals, who have traditionally faced barriers
to loans from commercial banks or have chosen
not to maintain checking or savings accounts
because of steep maintenance fees. Elektra, whose
retail operations include mostly consumer electronics and home appliances, for example, provided consumer loans for about five decades
before creating a commercial bank.
Wal-Mart entered the retail chain market in
Mexico in 1991 when the first Sam’s Club opened
in December. In 1992, Wal-Mart started an alliance
with Grupo Cifra, which controlled a successful
chain of retail stores. In 1993, Wal-Mart’s first
55

See Haber and Musacchio (2005).

See Bubel and Skelton (2005) and also Haber (2005b).

56

See Schulz (2004).

See Haber (2005c) and also Haber and Musacchio (2005).

57

See Bubel and Skelton (2005).

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

429

Hernández-Murillo

Supercenter opened, and more companies (a
chain of restaurants and a chain of clothing stores)
joined the alliance with Cifra. Wal-Mart quickly
gained market share and in 1997 bought a controlling stake in Grupo Cifra. By 2001, Wal-Mart
de México controlled about half of all supermarket sales in the country. In 2006 alone, 120 stores
were opened and total sales reached about $18.3
billion, an inflation-adjusted increase of 15.9
percent over 2005 sales.58 Today, Wal-Mart is
Mexico’s largest private employer, with 149,584
employees; it owns 917 stores in 147 cities, which
include supermarkets, restaurants, and clothing
stores.
Wal-Mart’s banking experiment is interesting
because of the regulatory environment it will face
in Mexico. Gelpern (2007) notes that all other
foreign banks in Mexico, although established by
law as wholly owned and separately capitalized
subsidiaries, have parent institutions that are
themselves regulated and subject to supervision by
their home country authorities. Given Wal-Mart’s
failed attempts to enter the banking sector in the
United States, Wal-Mart’s bank in Mexico remains
its only banking venture in the world, and therefore the Mexican authorities will have sole responsibility over its regulation and supervision, while
Wal-Mart’s headquarters in the United States will
remain outside of Mexican authorities’ regulatory
scope. Gelpern warns, however, that although
the banking license establishes limits to the ties
between Wal-Mart’s banking and retail businesses,
Mexico’s limited leverage over Wal-Mart activities
outside of Mexico puts Wal-Mart in a unique position to demand support from Mexican authorities.
This, Gelpern argues, is because of Wal-Mart’s
strength in the retail sector, because no other foreign bank has retail stores, and because other
retail chains with banks are currently owned by
Mexican nationals and therefore subject entirely
to Mexican laws.
Thus far, Wal-Mart’s strategy in entering the
banking sector appears to be similar to Elektra’s:
first, approach customers who have traditionally
avoided the banking sector; afterwards, expand
58

Source: sales and store counts from www.walmartmexico.com.mx;
United States–Mexico exchange rate as of December 2006 from
research.stlouisfed.org/fred2/data/EXMXUS.txt.

430

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

2007

to middle-income consumers. The effects of
Wal-Mart’s entry on banking competition are yet
to be seen. However, Wal-Mart’s entry into the
banking sector should provide for an interesting
environment in the very immediate future: Despite
the improved conditions in the sector, Mexican
consumers still face high fees for banking services, and Wal-Mart may extend its traditionally
aggressive low-pricing strategy to its banking
business.

WHAT NEXT?
Apart from recent favorable developments in
Mexico’s bank credit, at its core, the country still
faces fundamental problems with its legal infrastructure. Property rights in Mexico were until
very recently not well defined and in many cases
continue to be poorly enforced.59 Even today, it
is difficult even to ascertain who owns a particular asset (recall that a unified property registry
was not put in place until 2006); and, as a consequence, it is still difficult for banks to repossess
assets given as collateral in a loan. Haber (2005c)
argues that, as long as these institutional problems
are not resolved, they will continue to limit the
type of contracts that bankers can enforce. Haber’s
view reflects the notion that the protection of
property rights has positive effects on long-run
economic performance, a notion that is widely
recognized by economists. (See, for example,
Acemoglu and Robinson, 2000; Acemoglu,
Johnson, and Robinson, 2002; and, more recently,
Gradstein, 2007, among others.) It would seem,
then, that Mexico stands to gain from continuing
to foster the development of its banking sector
and, as a long-run goal for its overall development, modernize its legal infrastructure, especially contract enforcement and the definition
and defense of private property rights.

REFERENCES
Acemoglu, Daron; Johnson, Simon and Robinson,
James A. “Reversal of Fortune: Geography and
59

For an analysis of property rights in the Mexican Constitution,
see also Katz (1998).

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Hernández-Murillo

Institutions in the Making of the Modern World
Distribution.” Quarterly Journal of Economics,
November 2002, 117(4), pp. 1231-94.
Acemoglu, Daron and Robinson, James A. “Why Did
the West Extend the Franchise? Democracy,
Inequality, and Growth in Historical Perspective.”
Quarterly Journal of Economics, November 2000,
115(4), pp. 1167-99.
Avalos, Marcos and Hernández Trillo, Fausto.
“Competencia Bancaria en México.” Naciones
Unidas. Comisión Económica para América Latina
(CEPAL), November 2006, Serie Estudios y
Perspectivas 62. México.
Bubel, Robert V. and Skelton, Edward C. “Financial
Globalization: Manna or Menace? The Case of
Mexican Banking.” Federal Reserve Bank of Dallas
Southwest Economy, January/February 2002, 9(1),
pp. 17-19.
Bubel, Robert V. and Skelton, Edward C. “Mexico
Emerges from 10-Year Credit Slump.” Federal
Reserve Bank of Dallas Southwest Economy,
May/June 2005, 12(3), pp. 14-18.
Calvo, Guillermo A. and Mendoza, Enrique G.
“Mexico’s Balance-of-Payments Crisis: A Chronicle
of a Death Foretold.” Journal of International
Economics, November 1996, 41(3-4), pp. 235-64.
Cull, Robert; Senbet, Lemma W. and Sorge, Marco.
“Deposit Insurance and Financial Development.”
Journal of Money, Credit, and Banking, February
2005, 37(1), pp. 43-82.
Del Ángel, Gustavo A.; Haber, Stephen and
Musacchio, Aldo. “Normas Contables Bancarias en
México. Una Guía de los Cambios para Legos Diez
Años Después de la Crisis Bancaria de 1995.” El
Trimestre Económico, October-December 2006,
73(4), pp. 903-26.
Demirgüç-Kunt, Asli and Detragiache, Enrica. “Does
Deposit Insurance Increase Banking System
Stability? An Empirical Investigation.” Journal
of Monetary Economics, October 2002, 49(7),
pp. 1373-406.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Economist. “A Danish Model in Aztec Dress.”
January 6, 2007, 382(8510), pp. 62.
Furlong, Frederick T. and Keeley, Michael C. “Capital
Regulation and Bank Risk Taking: A Note.” Journal
of Banking and Finance, November 1989, 13(6),
pp. 883-91.
Gelos, R. Gaston and Werner, Alejandro M. “Financial
Liberalization, Credit Constraints, and Collateral:
Investment in the Mexican Manufacturing Sector.”
Journal of Development Economics, February 2002,
67(1), pp. 1-27.
Gelpern, Anna. “Wal-Mart Bank in Mexico: Money to
the Masses and the Home-Host Hole.” Connecticut
Law Review, May 2007, 39(4), pp. 1513-38.
Gil-Díaz, Francisco. “The Origin of Mexico’s 1994
Financial Crisis.” Cato Journal, Winter 1998, 17(3),
pp. 303-13.
Gonzalez-Anaya, José Antonio and Marrufo, Grecia M.
“Financial Market Performance in Mexico.” Center
for Research on Economic Development and Policy
Reform, Stanford University, 2001.
Gradstein, Mark. “Inequality, Democracy and the
Protection of Property Rights.” Economic Journal,
January 2007, 117(516), pp. 252-69.
Gruben, William C. and McComb, Robert P.
“Liberalization, Privatization, and Crash: Mexico’s
Banking System in the 1990s.” Federal Reserve
Bank of Dallas Economic Review, First Quarter
1997, pp. 21-30.
Gruben, William C. and McComb, Robert P.
“Privatization, Competition, and Supercompetition
in the Mexican Commercial Banking System.”
Journal of Banking and Finance, February 2003,
27(2), pp. 229-49.
Haber, Stephen. “Banking With and Without Deposit
Insurance: Mexico’s Banking Experiments 18842004.” Unpublished manuscript, Stanford
University, 2005a.
Haber, Stephen. “Mexico’s Experiments with Banking
Privatization and Liberalization, 1991-2003.”

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

2007

431

Hernández-Murillo

Journal of Banking and Finance, August-September
2005b, 29(8-9), pp. 2325-53.
Haber, Stephen. “Why Institutions Matter: Banking
and Economic Growth in Mexico.” Unpublished
manuscript, Stanford University, 2005c.
Haber, Stephen and Musacchio, Aldo. “Foreign
Banks and the Mexican Economy, 1997-2004.”
Unpublished manuscript, Stanford University, 2005.
Hernández Trillo, Fausto and López Escarpulli,
Omar. “La Banca en México, 1994-2000.”
Economía Mexicana, Nueva Época, 2001, 10(2),
pp. 363-90.
Katz, Isaac. El Sistema Financiero Mexicano: Motor
Del Desarollo Económico. México: Diana and
Centro de Investigación para el Desarrollo, A.C.,
1990.
Katz, Isaac. “Un Análisis Económico de la
Constitución.” Gaceta de Economía ITAM, Spring
1998, pp. 161-88.
Keeley, Michael C. “Deposit Insurance, Risk, and
Market Power in Banking.” American Economic
Review, December 1990, 80(5), pp. 1183-1200.
La Porta, Rafael; López de Silanes, Florencio and
Zamarripa, Guillermo. “Related Lending.” Quarterly
Journal of Economics, February 2003, 188(1),
pp. 231-68.
Mackey, Michael W. “Informe de Michael W. Mackey
en la Evaluación Integral de las Operaciones y
Funciones del Fondo Bancario de Protección al
Ahorro, FOBAPROA, y la Calidad de Supervisión
de los Programas del FOBAPROA de 1995 a 1998.”
Gaceta Parlametaria, Año II, Número 352,
September 13, 1999, Cámara de Diputados, H.
Congreso de la Unión, México.

McQuerry, Elizabeth. “The Banking Sector Rescue in
Mexico.” Federal Reserve Bank of Atlanta
Economic Review, Third Quarter 1999, pp. 14-29.
McQuerry, Elizabeth and Espinosa, Marco.
“Reviewing Mexico’s New Bank Accounting
Standards.” Federal Reserve Bank of Atlanta
Economics Update, January-March 1998.
Murillo, José Antonio. “La Banca en México:
Privatización, Crisis, y Reordenamiento.”
Unpublished manuscript, Banco de México, 2002.
Negrin, José Luis. “Mecanismos para Compartir
Información Crediticia. Evidencia Internacional y
la Experiencia Mexicana.” El Trimestre Económico,
July-September 2001, 68(3), pp. 405-65.
Pickering, Natalie. “The SOFOLES: Niche Lending or
New Lenders in the Mexican Mortgage Market?”
Unpublished manuscript, Harvard University,
2000.
Schulz, Heiner. “Foreign Banks in Mexico: New
Conquistadors or Agents of Change?” Unpublished
manuscript, University of Pennsylvania, 2004.
Skelton, Edward C. “Laying the Foundation for a
Mortgage Industry in Mexico.” Federal Reserve
Bank of Dallas Economic Letter, October 2006,
1(10), pp. 14-18.
Trigueros, Ignacio. “El Sistema Financiero Mexicano.”
In Luis Rubio and Arturo Fernández, eds., México
a la hora del cambio. México: Cal y Arena and
Centro de Investigación para el Desarrollo, A.C.,
1995.
Unal, Haluk and Navarro, Miguel. “The Technical
Process of Bank Privatization in Mexico.” Journal
of Financial Services Research, September 1999,
16(1), pp. 61-83.

Martin, Antoine. “Liquidity Provision vs. Deposit
Insurance: Preventing Bank Panics Without Moral
Hazard.” Economic Theory, 2006, 28(1),
pp. 197-211.

432

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

How Well Does Employment Predict Output?
Kevin L. Kliesen
Economists, policymakers, and financial market analysts typically pay close attention to aggregate
employment trends because employment is thought to be an important indicator of macroeconomic
conditions. One difficulty is that there are two separate surveys of employment, which can diverge
widely from one another, as the previous and current economic expansions demonstrate. The
conventional wisdom is that, for assessing economic conditions, the survey that counts the number
of jobs (establishment survey) is preferable to the survey that counts the number of people employed
(household survey). However, results from a one-quarter-ahead forecasting exercise presented in
this paper suggest that analysts should question whether employment is a useful indicator for
predicting output growth. (JEL C53, E24)
Federal Reserve Bank of St. Louis Review, September/October 2007, 89(5), pp. 433-46.

E

ach month, the Bureau of Labor
Statistics (BLS) publishes the
Employment Situation, a comprehensive report of key labor market statistics
derived from separate surveys of nonagricultural
business establishments (including government)
and private households. Labor market conditions
are among the most closely watched of all sources
of economic statistics because they are, collectively, thought to be a comprehensive indicator
of economic activity. In addition, many of its
components (for example, hours worked) are
used in the construction of other key economic
statistics, such as the Federal Reserve Board’s
index of industrial production (IP). As a result,
policymakers and economic analysts monitor
labor market conditions carefully to help them
gauge the evolving strength or weakness in the
pace of aggregate economic activity over the nearterm. In fact, policymakers and analysts use these
data, ideally, to more accurately predict economic
activity and avoid forecast surprises, which have

been shown to significantly affect the behavior
of financial markets.1
This article will first briefly discuss the two
surveys that the BLS uses to measure employment
and then discuss the different patterns in employment growth registered by each measure over the
past two business cycles. The article will provide
some evidence, with a one-quarter-ahead pseudo
out-of-sample forecasting exercise, that neither
measure of employment is a reliable predictor of
the growth of IP or real gross domestic product
(GDP).

TWO EMPLOYMENT SURVEYS
One difficulty that confronts economic
analysts who use employment to help predict
the growth of real GDP is which employment
measure to use. The BLS publishes two measures
of employment that receive the most attention.
1

See Kliesen and Schmid (2004 and 2006).

Kevin L. Kliesen is an associate economist at the Federal Reserve Bank of St. Louis. The author thanks Bill Emmons, Mike McCracken, and
Dan Thornton for helpful suggestions and Joshua A. Byrge for research assistance.

© 2007, The Federal Reserve Bank of St. Louis. Articles may be reprinted, reproduced, published, distributed, displayed, and transmitted in
their entirety if copyright notice, author name(s), and full citation are included. Abstracts, synopses, and other derivative works may be made
only with prior written permission of the Federal Reserve Bank of St. Louis.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

433

Kliesen

Table 1
Differences Between the CES and the CPS
CPS

CES
• Monthly sample survey of 160,000 businesses and
government agencies; firms of all sizes are included

• Monthly sample survey of approximately 60,000
households

• Designed to measure employment, hours, and
earnings with significant industrial and geographic
detail

• Designed to measure employment and
unemployment with significant demographic detail

• Reference period is the pay period (could be weekly,
biweekly, monthly, and so forth) that includes the
12th of the month

• Reference period is the week that includes the 12th
of the month

• Employees of all ages are included

• Only workers aged 16 and older are included

• Employment measure reflects the number of
nonfarm payroll jobs

• Employment measure reflects the number of
employed persons

• Multiple jobholders are counted for each payroll job

• Multiple jobholders are counted once

• Self-employed persons are excluded

• Self-employed persons are included

• Agriculture sector is excluded

• Agriculture sector is included

• Private household workers (nannies, housekeepers,
and the like) are excluded

• Private household workers are included

• Unpaid family workers (persons working without
formal pay in their family’s business are excluded

• Unpaid family workers are included

• Workers on leave without pay throughout the
reference period are excluded

• Workers on leave without pay throughout the
reference period are included

NOTE: Reprinted from Bowler and Morisi (2006, p. 24) with permission.

One is nonfarm payroll employment. This measure
is taken from the Current Employment Statistics
survey (CES), which is a survey of about 160,000
businesses and government agencies (establishments) that cover about a third of all nonfarm
payroll workers. The other measure is civilian
employment, which is taken from the Current
Population Survey (CPS). The CPS is a survey
of about 60,000 households, of roughly 110 million households in the United States. It is often
referred to as the household survey or household
employment.
These two series are reported each month in
the Employment Situation, which is typically
released on the first Friday of each month by the
BLS. According to the BLS, the CES is designed
chiefly to track changes such as jobs and hours,
whereas the CPS is designed chiefly to track
changes in rates, such as the unemployment rate
or the labor force participation rate.2
434

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

2007

As seen in Table 1, there are several differences
between the two surveys. For example, the CPS
(household survey) includes agricultural and
unincorporated self-employed workers, as well
as those who are on unpaid absence; the CES
(establishment survey) includes none of these.3
Also, the CES includes multiple jobholders and
workers of all ages, whereas the CPS counts an
employed person only once and only those who
are at least 16 years old. Finally, each year the CES
estimates are benchmarked to the actual employment levels reported by the state unemployment
2

Bowler and Morisi (2006).

3

There are two measures of self-employment: incorporated and
unincorporated. Unincorporated self-employed individuals have
the status of unpaid family workers—that is, they are assumed
not to be paid employees. These are reported in the CPS only.
However, the incorporated self-employed are assumed to be paid
employees of their own corporations and, hence, are counted in
the CES.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Kliesen

insurance records.4 Typically, the benchmark
revisions are relatively small. According to the
BLS, the benchmark revisions averaged ± 0.2 percent from 1996 to 2005. However, the 2006 benchmark revision was unusually large, measuring 0.6
percent.5 In contrast, the CPS estimates are derived
from annual estimates of the civilian noninstitutional population provided by the Bureau of the
Census.6
Although the coverage of the labor market is
broader in the household survey, most economists
probably place more emphasis on the establishment survey because it is constructed from a
much larger sample and is less volatile than the
household survey. Of course, this assumes that
each survey is a statistically representative sampling of the population; typically, a larger sample
of the population produces more accurate estimates than a smaller sample of the population.
To see why the CES might be a superior timeseries measure, consider the following. The BLS
survey of businesses and government agencies
encompasses about 400,000 individual worksites,
which covers about a third of all workers.7 Thus,
the BLS is implicitly surveying a little more than
45 million workers. By contrast, the CPS encompasses about 60,000 households (or about 76,000
workers). This represents only 0.05 percent of
total households and 0.2 percent of those implicitly surveyed by the CES.8,9
4

These data are reported every three months in the Quarterly
Census of Employment and Wages, also known as the ES-202. See
www.bls.gov/cew/home.htm.

5

See www.bls.gov/web/cesbmart.htm.

6

The population controls for the CPS are based on the decennial
census. Intercensal population estimates from the Census Bureau
are based on births, deaths, and estimates of net internal
immigration.

7

See BLS Handbook of Methods (Chapter 2);
www.bls.gov/opub/hom/homch2_a.htm.

8

According to Bowler and Morisi (2006), the monthly CPS sampling
error is four times larger than that of the CES.

9

The CPS surveys about 60,000 households. This represents only
0.05 percent of total households. Viewed another way, there are
1.26 employees per household, which represents nearly 76,000
persons surveyed. This sample is less than 0.2 percent of those
implicitly surveyed by the CES.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Reconciling the Two Employment
Measures
Despite the marked differences in their structural characteristics, the two series have increased
at comparable rates since their inception. From
1948 to 2006, the CES has grown at a 1.6 percent
annual rate, whereas the CPS has grown at a 1.9
percent annual rate. At shorter time horizons,
though, the growth rates have diverged considerably. For example, Figure 1 shows that peak and
trough growth rates of the CES are generally larger
(in absolute terms) than those of the CPS. To help
reconcile divergences, the BLS creates an adjusted
series that attempts to net out, as much as possible, the methodological differences between the
two surveys.
Each month, the BLS computes an employment measure from the CPS that is designed to
be conceptually equivalent to that of the CES.
This adjusted measure has been published on a
monthly basis since January 1994, when there was
a major redesign of the CPS.10 To make the CPS
measure conceptually equivalent to the CES measure, as seen in Table 2, the BLS subtracts the following employment categories of the CES that are
outside of the scope of the CPS: (i) agriculture,
(ii) nonagricultural self-employment, (iii) nonagricultural unpaid family workers, (iv) private
household workers, and (v) unpaid absences. To
this adjustment, the BLS also adds the number
of workers with multiple jobs.11 Table 2 shows,
as of 2006, that CES employment totaled 136.935
million and CPS-adjusted employment totaled
139.415 million—an unexplained difference of
2.48 million. (In December 2005, it was 1.67
million.)
Figure 2 shows the evolution of the CES-CPS
discrepancy during the past two business cycles.
Panel A shows employment trends during the
1990s expansion, and panel B shows employment
trends during the current business expansion.
Each employment series is indexed to equal 1
during the trough months of the recessions in
10

See Polivka and Miller (1998).

11

The number of multiple jobholders are those 16 and older with
secondary jobs.

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

2007

435

Kliesen

Figure 1
Employment Growth, 1950 to 2006
Four-Quarter Percent Changes

CES

10

CPS
8
6
4
2
0
–2
–4
–6
1950

1955

1960

1965

1970

1975

1980

1985

1990

1995

2000

2005

SOURCE: BLS and author’s calculations.

Table 2
CES-CPS Reconciliation Table (in thousands)
Employment

December 2005

December 2006

Percent change

Payroll jobs

135,041

136,935

1.4

Household employment*
Less:

142,918

146,081

2.2

Agriculture

1,942

2,079

7.1

Nonagricultural self-employed

9,294

9,751

4.9

66

99

50.0

Nonagricultural unpaid family workers
Private household workers
Unpaid absences
Total
Plus:

Multiple jobholders†

Adjusted household employment*
Adjusted household employment less payroll jobs
Addenda: Nonagricultural wage and salary workers

782

722

–7.7

1,412

1,547

9.6

13,496

14,198

5.2

7,289

7,533

3.3

136,711

139,415

2.0

1,670

2,480

48.5

130,755

132,901

1.6

NOTE: *The changes in household employment and adjusted household employment have been adjusted to account for the introduction
of new population controls in January 2006. †Multiple jobholders who are nonagricultural wage and salary workers on their primary
job. Totals and changes in household employment series are based on unrounded numbers. Data are not seasonally adjusted, and the
published numbers are those prior to the March 2006 benchmark revisions published in February 2007.
SOURCE: BLS.

436

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Kliesen

Figure 2
Establishment and Household Employment During the 1991-2001 and Current
Business Expansion
A. January 1990 to December 2000 (Index, March 1991 = 1.0)
1.25
1.20
1.15
1.10
CES

1.05

CPS
Trough

1.00
0.95
Jan 90

Jul 91

Jan 93

Jul 94

Jan 96

Jul 97

Jan 99

Jul 00

B. January 2001 to December 2006 (Index, November 2001 = 1.0)
1.08
1.07
1.06
1.05
1.04
1.03
1.02
CES

1.01

CPS

1.00

CPS Adjusted

0.99

Trough

0.98
Jan 01

Jan 02

Jan 03

Jan 04

Jan 05

Jan 06

NOTE: Data are seasonally adjusted by the BLS and reflect the March 2006 benchmark revision published in February 2007. Vertical
lines indicate the month declared to be the end of the recession (trough).
SOURCE: BLS and author’s calculations.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

437

Kliesen

1990-91 (July 1990) and 2001 (November 2001),
respectively. The CES and CPS measures are
shown in each panel, but panel B also includes
the CPS-adjusted measure. Recall that this series
has been available only since January 1994.12
Although the CPS growth and CPS-adjusted
rates have kept up with the CES growth rate overall since 1994, the CES and CPS measures have
moved in distinct patterns since that year (see
Figure 2A).13 From January 1994 to March 2001,
increases in payroll employment (CES) averaged
about 233,000 per month, while they averaged
only about 184,000 per month for household
employment (CPS). From November 2001 to
December 2006, the opposite occurred: Household
employment increased by an average of about
159,000 per month, while payroll employment
increased by an average of only 103,000 per
month. Panel B also shows that the CPS-adjusted
measure has increased in line with total household employment, although there has been some
slight widening between the two measures in
2006. From November 2001 to December 2006,
the CPS-adjusted measure increased by 155,000
per month, which is virtually identical to the
average increase in the unadjusted CPS.

Analysis in the Literature
Numerous studies have attempted to explain
the discrepancies between the two growth rates
of the two main employment series during these
two periods. Many researchers have focused on
structural changes in the economy, such as innovations in the production and distribution of
manufactured goods, or increases in trend labor
productivity growth rates: See Kahn (1993), Schreft
and Singh (2003), and Groshen and Potter (2003).
In this vein, Koenders and Rogerson (2005) argue
that slow employment growth tends to follow
long expansions—such as those in 1982-90 and
1991-2001—because firms postpone structural
changes during periods of relatively strong growth.
12

The CPS-adjusted series is adjusted to smooth out the discrete
jumps in population that occurred because of revisions in January
2000, 2003, 2004, 2005, and 2006.

13

From January 1994 to December 2006, the average monthly gains
are as follows: CES, 159,000; CPS, 155,000; CPS adjusted, 159,000.

438

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

2007

One potential difficulty with these explanations,
as Aaronson, Rissman, and Sullivan (2004) point
out, is that structural change is a regular feature
of a dynamic economy that is continually subject
to waves of creative destruction.
Bowler and Morisi (2006) take a somewhat
different approach by reviewing several possible
explanations tied to methodological differences
between the two employment surveys. They found
that the population undercount explained roughly
one-third of the 4.5 million cumulative gap
between the CES and CPS measures of employment that arose in the late 1990s.14 This is consistent with an earlier finding by Juhn and Potter
(1999), who argued that the widening gap between
the CES and CPS in the 1990s was likely due to
an underestimate of the working-age population.
Recall that employment estimates in the CPS are
derived from estimates of the noninstitutional
population that are provided by the Bureau of
the Census. In January 2003, the BLS incorporated
into the CPS the decennial census population
estimate. This introduction resulted in revisions
to the population controls, and thus the CPS data,
from January 2000 through December 2002.15 A
second major revision occurred in January 2003
because of the annual intercensal population
adjustment.16
Bowler and Morisi also reviewed other
sources of systematic error, such as workers with
multiple jobs, military personnel with secondary
civilian jobs, foreign commuters, and differences
in the reference periods between the two surveys.
None of these factors, they argued, could explain
the 1990s discrepancy with any significance.
Regarding developments since 2001—that is, the
relatively faster growth of household employment—they argue that the causes of this discrepancy are also not fully known. Two possibilities
are increased job turnover and potential inaccuracies in the population controls. An earlier BLS
paper by Nardone et al. (2003) argued that undoc14

Their sample period was 1994-2000. The 4.5 million gap does not
include the population revisions associated with the introduction
of the Census 2000 population controls in January 2003.

15

See Bowler et al. (2003).

16

See footnote 4.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Kliesen

Table 3
Correlation Between Various Measures of Employment and Output Growth (one-quarter
percent changes)
1950:Q1 to 1993:Q4

1994:Q1 to 2001:Q4

2002:Q1 to 2006:Q4

CES

CPS

CES

CPS

CES

CPS

IP

0.82

0.63

0.89

0.62

0.28

0.08

Real GDP

0.72

0.56

0.59

0.37

0.24

–0.02

SOURCE: Author’s calculations.

umented immigration may also bias population
controls and thus also the estimates of CPS
employment. The divergent pattern of employment growth in the current expansion potentially
presents another difficulty for economic analysts
who use employment as a predictor of near-term
economic activity.

Reserve Chairman Ben Bernanke (2003), is that
the CES is the preferred measure:

GAUGING THE INFORMATION
CONTENT OF THE EMPLOYMENT
MEASURES

Table 3 provides a preliminary assessment of
whether economic analysts should continue to
rely more on the CES or put more weight on the
CPS. Table 3 shows simple correlations between
the annualized one-quarter growth rate of two
measures of economic activity—IP and real GDP—
and the two primary measures of employment
plotted in panel A of Figure 2—CES and CPS.
The table lists the correlations over three separate
periods: (i) 1950:Q1–1993:Q4; (ii) 1994:Q1–
2001:Q4; and (iii) 2002:Q1–2006:Q4. The 1994
breakpoint is chosen because, as noted earlier,
the CPS was changed in several important ways
in that year; the 2001 break point was chosen to
be consistent with the analysis presented earlier.
As seen in Table 3, correlations between the
growth of employment and IP are invariably
stronger than those between the growth of employment and real GDP. This result holds across all
three periods and for each employment series.
This finding seems surprising given that IP measures the output of the goods-producing sector,
which is only about 35 percent of real GDP.17 It

In the short term, changes in the demand for
goods and services are assumed to cause changes
in the demand for labor. But because the employment statistics are reported in advance of the
primary measure of aggregate output (real GDP),
policymakers tend to view them as a gauge for
future changes in output. However, predicting
output based solely on changes in labor input is
complicated by changes in labor productivity.
A key question among economic analysts
during the 1990s was whether the faster growth
of CES employment was a more accurate portrayal
of underlying trends in employment than was
household employment. With the converse being
true since the 2001 recession, a parallel question
is whether the relatively faster growth of household employment is a truer measure of the economy’s employment growth and thus its near-term
growth of economic activity. If so, should economic analysts pay more attention to CPS employment to predict real GDP growth? Or should they
continue to rely more on the CES? The conventional wisdom, as recently expressed by Federal
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

[B]ecause of the larger sample used in the
payroll survey and because of possible problems with the population estimates used to
scale the household survey, somewhat greater
reliance should probably be placed on the payroll survey.

17

This share increases to 45 percent if structures are added to goods
production. In nominal terms, the goods share is 31 percent and
the sum of goods and structures is 42 percent.

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

2007

439

Kliesen

is also the case that the correlation coefficients
for the CES series are larger than those for the
CPS series. This is consistent with the conventional wisdom noted earlier that most economists
follow: The CES is as a better measure of employment conditions.
A second finding from Table 3 is that the correlations between employment and output have
changed significantly over time.18 For example,
the correlation coefficient between the growth
of CES employment and IP increased from 0.82
in the 1950-93 period to 0.89 in the 1994-2001
period. This result may reflect both the faster
growth of CES employment during the 1990s
and the strong gains in labor productivity and
equity prices that helped spur a boom in business
capital spending. Since 2001, the correlation
between the growth of CES employment and IP
has dropped to 0.28. A similar finding is found
for real GDP growth and employment. Over time,
the correlation between the growth of CES employment and real GDP declined from 0.72 (1950-93)
to 0.59 (1994-2001) to 0.24 (2002-06). A similar
result was found for the correlation between the
growth of CPS employment and real GDP growth.
In fact, the CPS-real GDP correlation during the
current period (2002-06) is now negative. This
seems surprising given that the growth of CPS
employment has been stronger than CES employment over this period.
The finding that the correlation between the
growth of household employment and real GDP
has decreased over time contrasts with the results
shown in Perry (2005). He provides evidence that,
since 1994, household employment has been a
better measure of aggregate employment than
payroll employment when predicting real GDP
growth or growth of nonfarm business output (a
subset of GDP). Perry concludes that both measures include important information and that an
economic analyst might benefit from adopting
18

The correlations in Table 3 are somewhat sensitive to the end
points, depending on whether the 2001 recession is moved to the
second or third period. For example, choosing 2000:Q4 as the end
point of the second period lowers the correlations between employment and output in the second period while raising them in the
third period. However, this does not change the overall conclusion
that the correlations between employment and output growth have
generally weakened over time.

440

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

2007

the simple rule of thumb of averaging the monthly
changes in the two series. However, the correlations presented in Table 3 suggest that economic
analysts should pay less attention to the CPS relative to the CES (conventional wisdom). In the
next section, I formally assess the validity of this
conventional wisdom.

FORECASTING OUTPUT USING
MEASURES OF EMPLOYMENT
Expectations about the state of the U.S. economy matter tremendously to policymakers.
Accordingly, the Federal Open Market Committee
(FOMC) closely monitors the incoming flow of
economic data to see whether its information
content is consistent with the prevailing forecast
for inflation and key indicators of real activity.
Chairman Bernanke and other senior Federal
Reserve officials have called this a “datadependent policy.”19 One of the most important
data series that shapes the FOMC’s view is the
monthly change in employment. The purpose
of this section is to assess whether employment
changes are a good predictor of real output growth
over the following quarter.
For this analysis, I use two measures of output
and six measures of employment, with data measured at a quarterly frequency. The two measures
of output are real GDP and IP, the same measures
used in Table 3. There are six measures of
employment:
• nonfarm payroll employment (CES);
• civilian employment (CPS);
• civilian wage and salary workers (CPS
W&S);
• the Perry (2005) rule-of-thumb series,
which measures the average change of the
CES and CPS (CES-CPS AVG);
• nonfarm payroll employment plus the measure of self-employed workers (CES + SE);
• the CPS less the measure of self-employed
workers from the CPS (CPS – SE).
19

See Poole (2006).

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Kliesen

Table 4
Summary Statistics for Output and Employment (one-quarter percent changes, not annualized)
Real GDP

IP

CES

CPS

CPS W&S

CES-CPS
AVG

CES + SE

CPS – SE

Average

0.89

0.94

0.54

0.42

0.52

0.48

0.51

0.45

Standard deviation

1.09

2.24

0.70

0.56

0.63

0.59

0.65

0.61

Average

0.80

0.83

0.39

0.36

0.40

0.37

0.37

0.37

Standard deviation

0.48

0.93

0.35

0.36

0.41

0.31

0.35

0.36

1950:Q1 to 1993:Q4

1994:Q1 to 2006:Q4

SOURCE: Author’s calculations.

The rationale for including the final two
measures stems from the fact that the growth of
self-employment has been considerably stronger
in the period following the 2001 recession.20
From November 2001 through December 2006,
self-employment increased by 11 percent, while
total CPS employment increased by 7.1 percent.
Hence, self-employment may be one potential
explanation for the current growth gap between
the CES and the CPS. Finally, I do not use the
BLS’s CPS series that is conceptually equivalent
to the CES because that series is not available
before the first quarter of 1994.
Table 4 provides some basic statistics for the
economic series that will be used in the empirical
analysis. I look at two periods: 1950:Q1–1993:Q4
and 1994:Q1–2006:Q4. Table 4 offers a few key
findings. First, average growth of real GDP and IP
are approximately equal in each period. Second,
volatility declined by more than half for GDP
growth and by slightly more than that for IP
growth in the second period. Third, the growth
of CES employment is larger than the growth of
CPS employment, but the gap narrowed by threefourths in the second period. Finally, volatility
has also declined for employment growth, though
less than it has for output. Moreover, in the second
period, CES and CPS employment volatility are
20

From 1948 to 1966, self-employment averaged 9.5 percent of
civilian employment, but it has held fairly steady since the mid1960s, fluctuating between 6.5 and 7.5 percent since 1967.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

approximately equal, which contrasts with the
earlier period. Although, the differences in volatility among the employment series are small, the
CES-CPS average is the least volatile.
I use the following autoregressive model (AR),
with four lags of the dependent variable, as the
benchmark one-quarter-ahead forecast for the
quarterly change in output:
4

(1)

X t = α t + ∑ βi X t − i + µt ,
i =1

where Xt is the log change (annualized) in real
GDP or IP. To test whether adding information
from employment produces a more accurate onequarter-ahead forecast than the benchmark model,
I add, in six separate regressions, the log changes
of employment described earlier. Finally, I consider two separate modifications to the benchmark
model. In the first specification, I add the first
lag of the quarterly log change (annualized) in
employment, Yt –1, to the baseline model:
4

(2)

X t = α t + ∑ βi X t − i + γ Yt −1 + µt .
i =1

In the second specification, I add (separately)
the contemporaneous value of the quarterly log
change (annualized) in employment, Yt , to the
baseline model:
4

(3)

X t = α t + ∑ βi X t − i + γ Yt + µt .
i =1

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

2007

441

Kliesen

This form of a nested, one-step-ahead forecast is
common in the literature, where (1) is termed the
restricted model and (2) and (3) are unrestricted
models.21
Tables 5 and 6 show the root mean-squared
errors (RMSE) from this one-step-ahead pseudo
out-of-sample forecasting exercise for the restricted
model and for both specifications of the unrestricted model.22 In both tables, the starting point
for the estimation period is the first quarter of
1950. To see how the one-step-ahead pseudo
out-of-sample forecasting exercise is conducted,
consider Table 5. In this case, the out-of-sample
forecast horizon is from 1994:Q1 to 2001:Q4.
Thus, the model is estimated using data from
1950:Q1 to 1993:Q4. Next, the model forecasts
the growth of real GDP in 1994:Q1 and uses either
the lagged (specification (1)) or contemporaneous
(specification (2)) growth of employment in
1994:Q1. To forecast the growth of real GDP in
1994:Q2, the model is run using actual data
through 1994:Q1 and so forth for each quarter.
In essence, the forecasting exercise tests whether
knowing the quarterly change in employment—
either the contemporaneous change or its value
in the previous quarter—is useful for predicting
the growth rate of real GDP or IP.23 A similar exercise is repeated in Table 6, but the out-of-sample
forecast horizon is from 2002:Q1 to 2006:Q4.
RMSEs in Tables 5 and 6, then, are the averages
for the out-of-sample forecast period indicated
for each model.
Tables 5 and 6 indicate that the RMSEs from
the restricted and unrestricted models are always
less for forecasts of real GDP than for IP. This perhaps reflects that (i) more than half of GDP is services output and (ii) goods output is more volatile
than services output. Another key finding is that
the RMSEs for the benchmark forecasts for IP and
21

See Clark and McCracken (2001), who cite the seminal work of
Meese and Rogoff (1983). See Stock and Watson (2003) for an
application to forecasting inflation with asset prices.

22

The forecasts are termed “pseudo” because actual data are used to
generate a one-step-ahead forecast that will be compared with the
actual growth rate.

23

The data in this article incorporate all previous revisions (currentvintage data). An interesting extension of this exercise would be a
test of whether the results would change if real-time data were used.

442

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

2007

real GDP growth are significantly lower in the
latter period (Table 6) than in the earlier period
(Table 5). This finding suggests that reduced
volatility helps a forecaster produce more accurate (lower RMSEs) forecasts.24
A more interesting finding from the tables is
that forecast accuracy in either period appears to
depend significantly on whether the lagged or
contemporaneous value of employment is used
to forecast output. Table 5 suggests that adding
the contemporaneous value of CES employment
substantially lowers the RMSE in the earlier period
but not in the latter period. For example, in the
forecast period encompassing 1994:Q1–2001:Q4,
adding CES employment lowers the RMSE for the
IP forecast from 3.25 percent to 2.48 percent and
for real GDP from 2.05 percent to 1.65 percent.
Using the lagged value of the CES in the earlier
period lowers the RMSE for real GDP only slightly
(from 2.05 to 2.00), but it raises the RMSE slightly
for IP (from 3.25 to 3.31). The remaining specifications of the unrestricted model in the earlier
period uniformly produced higher RMSEs than
the contemporaneous value of the CES.
In contrast, Table 6 shows the opposite finding. In the latter period (2002:Q1–2006:Q4),
adding the lagged value of the CES to the restricted
model produces a lower RMSE, whereas using
the contemporaneous value of the CES produces
a higher RMSE. For example, using the lagged
specification, adding the CES to the restricted
model lowers the RMSE for IP from 2.78 to 2.50
percent; for real GDP, the RMSE declines from
1.61 to 1.54 percent. A similar-size RMSE is produced using either the CPS W&S or the CES-CPS
AVG in the real GDP forecast in the lagged specification. The latter finding is consistent with
Perry’s (2005) results noted earlier. Adding the
self-employed to the CES (CES + SE) produces a
more accurate forecast than the contemporaneous
value of the CES by itself; however, the result
fails to hold when using the lagged values of
employment in this period.
24

This point is made by Stock and Watson (2005). They argue that
RMSEs for inflation and output forecasts have been reduced since
the mid-1980s because of the Great Moderation. Regarding the
latter, see McConnell and Pérez-Quirós (2000) or Ahmed, Levin,
and Wilson (2004).

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Kliesen

Table 5
One-Quarter-Ahead Forecast Errors (RMSEs), 1994:Q1 to 2001:Q4
Real GDP

IP
Lagged

Contemporaneous

Lagged

Contemporaneous

3.25

3.25

2.05

2.05

CES

3.31

2.48

2.00

1.65

CPS

3.24

3.33

2.01

2.18

CPS W&S

3.42

3.62

2.07

2.22

CES-CPS AVG

3.27

2.94

1.99

1.90

CES + SE

3.27

2.85

1.91

1.97

CPS – SE

3.27

3.14

2.06

1.97

Restricted model
AR(4)
Unrestricted models

NOTE: CES, nonfarm payroll employment; CPS, civilian employment; CPS W&S, civilian employment, nonfarm wage, and salary workers;
CES-CPS AVG, average of civilian and nonfarm payroll employment; CES + SE, nonfarm payroll employment plus self employment;
CPS – SE, civilian employment less self employment. First actual observation: 1950:Q1.

Table 6
One-Quarter-Ahead Forecast Errors (RMSEs), 2002:Q1 to 2006:Q4
IP

Real GDP

Lagged

Contemporaneous

Lagged

Contemporaneous

2.78

2.78

1.61

1.61

2.50

3.24

1.54

1.87

Restricted model
AR(4)
Unrestricted models
CES
CPS

2.75

3.02

1.57

1.81

CPS W&S

2.77

3.19

1.53

2.06

CES-CPS AVG

2.56

2.74

1.54

1.77

CES + SE

2.55

2.96

1.61

1.75

CPS – SE

2.73

3.08

1.56

1.91

NOTE: See Table 5 for descriptions of the mnemonics.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

443

Kliesen

Table 7
Clark-McCracken Tests of Forecast Accuracy, 1994:Q1 to 2001:Q4
Real GDP

IP
Lagged

Contemporaneous

Lagged

Contemporaneous

–0.0004

0.0262

0.0012

0.0182

Unrestricted models
CES
CPS

0.0006

0.0095

0.0010

0.0046

–0.0002

0.0082

0.0003

0.0058

CES-CPS AVG

0.0002

0.0191

0.0015

0.0116

CES + SE

0.0000

0.0179

0.0031

0.0084

CPS – SE

0.0002

0.0109

0.0001

0.0084

CPS W&S

10% critical value: 0.473
NOTE: See Table 5 for descriptions of the mnemonics. In each case, the forecast from the unrestricted model is tested against the forecast
from the restricted model. The null is that there is on difference between the two forecasts.

Table 8
Clark-McCracken Tests of Forecast Accuracy, 2002:Q1 to 2006:Q4
IP

Real GDP

Lagged

Contemporaneous

Lagged

Contemporaneous

0.0070

0.0316

0.0056

0.0146

Unrestricted models
CES
CPS

0.0012

0.0038

0.0018

0.0018

CPS W&S

0.0025

0.0080

0.0043

0.0021

CES-CPS AVG

0.0057

0.0244

0.0044

0.0093

CES + SE

0.0059

0.0294

0.0034

0.0140

CPS – SE

0.0016

0.0057

0.0019

0.0018

10% critical value: 0.335
NOTE: See Table 5 for descriptions of the mnemonics. In each case, the forecast from the unrestricted model is tested against the forecast
from the restricted model. The null is that there is on difference between the two forecasts.

444

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Kliesen

Ultimately, the value of any forecast is its
accuracy. A standard test of forecast accuracy is
the nonparametric test proposed by Diebold and
Mariano (1995). However, as Clark and McCracken
(2001) point out, the Diebold and Mariano test is
not appropriate for nested models like the one
used in this paper. The reason is that the limiting
distribution of the Diebold and Mariano test is
not normal when the null hypothesis is equal predictive power (of the restricted and unrestricted
test).25 Tables 7 and 8 show test statistics based
on an alternative test proposed by Clark and
McCracken (2005). In each case, the forecasts
from the unrestricted models in Tables 5 and 6
are tested against the restricted model. The null
hypothesis is that the two forecasts have the same
predictive power.
Despite the sizable difference between the
RMSEs of the restricted model and certain unrestricted models—especially the latter featuring
the CES—the Clark-McCracken test statistics indicate that the information contained in quarterly
changes in employment does not significantly
improve upon the benchmark AR(4) forecast.
This result holds for both periods and regardless
of whether one uses the contemporaneous specification or the lagged specification. Accordingly,
employment does not appear to be a statistically
useful predictor of output growth in the following
quarter, which seems contrary to the conventional
wisdom.

jobs (establishment survey) is preferable to the
survey that counts the number of people employed
(household survey) when attempting to discern
current economic conditions. However, results
from the pseudo out-of-sample forecasting exercise presented in this paper suggest that analysts
question whether employment is a useful predictor of output growth over a one-quarter horizon.

REFERENCES
Aaronson, Daniel; Rissman, Ellen R. and Sullivan,
Daniel G. “Can Sectoral Reallocation Explain the
Jobless Recovery?” Federal Reserve Bank of Chicago
Economic Perspectives, Second Quarter 2004,
28(2), pp. 36-49.
Ahmed, Shaghil; Levin, Andrew and Wilson, Beth
Anne. “Recent U.S. Macroeconomic Stability:
Good Policies, Good Practices, or Good Luck?”
Review of Economics and Statistics, August 2004,
86(3), pp. 824-32.
Bernanke, Ben S. “The Jobless Recovery.” Presented
at the Carnegie Mellon University conference
Global Economic and Investment Outlook,
Pittsburgh, Pennsylvania, November 6, 2003.
Bowler, Mary and Morisi, Teresa L. “Understanding
the Employment Measures from the CPS and CES
Survey.” Monthly Labor Review, February 2006,
129(2), pp. 23-38; www.bls.gov/opub/mlr/2006/02/
art2full.pdf.

CONCLUSION
Economists, policymakers, and financial
market analysts typically pay close attention to
aggregate employment trends, because employment is thought to be an important indicator of
macroeconomic conditions. One difficulty is that
there are two separate surveys of employment,
and, moreover, these measures can diverge from
one another in a significant fashion, as the previous and current economic expansions demonstrate. The conventional wisdom is that the
employment survey that counts the number of
25

Also see McGuckin, Ozyildirim, and Zarnowitz (2007).

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Bowler, Mary; Ilg, Randy E.; Miller, Stephen;
Robison, Ed and Polivka. Anne. “Revisions to the
Current Population Survey Effective in January
2003.” Bureau of Labor Statistics, February 2003;
www.bls.gov/cps/rvcps03.pdf.
Clark, Todd E. and McCracken, Michael W. “Evaluating
Direct Multistep Forecasts.” Econometric Reviews,
October 2005, 24(4), pp. 369-404.
Clark, Todd E. and McCracken, Michael W. “Tests of
Equal Forecast Accuracy and Encompassing for
Nested Models.” Journal of Econometrics,
November 2001, 105(1), pp. 85-110.

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

2007

445

Kliesen

Diebold, Francis X. and Mariano, Roberto S.
“Comparing Predictive Accuracy.” Journal of
Business and Economic Statistics, July 1995, 13(3),
pp. 253-63.

Meese, Richard A. and Rogoff, Kenneth S. “Empirical
Exchange Rate Models of the Seventies: Do They
Fit Out of Sample?” Journal of International
Economics, February 1983, 14(1-2), pp. 3-24.

Groshen, Erica L. and Potter, Simon M. “Has
Structural Change Contributed to the Jobless
Recovery?” Federal Reserve Bank of New York
Current Issues in Economics and Finance, August
2003, 9(8).

Nardone, Thomas; Bowler, Mary; Kropf, Jurgen;
Kirkland, Katie and Wetrogan, Signe. “Examining
the Discrepancy in Employment Growth between
the CPS and the CES.” Prepared for the Federal
Economic Statistics Advisory Committee,
October 17, 2003.

Juhn, Chinhui and Potter, Simon M. “Explaining the
Recent Divergence in Payroll and Household
Employment Growth.” Federal Reserve Bank of
New York Current Issues in Economics and Finance,
December 1999, 5(16).
Kahn, George A. “Sluggish Job Growth: Is Rising
Productivity or an Anemic Recovery To Blame?”
Federal Reserve Bank of Kansas City Economic
Review, Third Quarter 1993, 73(3), pp. 5-25.
Kliesen, Kevin L. and Schmid, Frank A.
“Macroeconomic News and Real Interest Rates.”
Federal Reserve Bank of St. Louis Review, March/
April 2006, 88(2), pp. 133-44.
Kliesen, Kevin L. and Schmid, Frank A. “Monetary
Policy Actions, Macroeconomic Data Releases, and
Inflation Expectations.” Federal Reserve Bank of
St. Louis Review, May/June 2004, 86(3), pp. 9-21.
Koenders, Kathryn and Rogerson, Richard.
“Organizational Dynamics Over the Business
Cycle: A View on Jobless Recoveries.” Federal
Reserve Bank of St. Louis Review, July/August
2005, 87(4), pp. 555-79.
McConnell, Margaret M. and Pérez-Quirós, Gabriel.
“Output Fluctuations in the United States: What
Has Changed since the Early 1980’s?” American
Economic Review, December 2000, 90(5),
pp. 1464-76.

Perry, George L. “Gauging Employment: Is the
Professional Wisdom Wrong?” Brookings Papers
on Economic Activity, 2005, (2), pp. 285-311.
Poole, William. “Data Dependence.” Presented at
the Middle Tennessee State University Annual
Economic Outlook Conference, Murfreesboro,
Tennessee, September 29, 2006.
Polivka, Anne E. and Miller, Stephen M. “The CPS
after the Redesign: Refocusing the Economic Lens,”
in John Haltiwanger, Marilyn E. Manser, and Robert
Topel, eds., Labor Statistics Measurement Issues:
Studies in Income and Wealth. Volume 60. Chicago:
University of Chicago Press, 1998, pp. 249-86.
Schreft, Stacey L. and Singh, Aarti. “A Closer Look
at Jobless Recoveries.” Federal Reserve Bank of
Kansas City Economic Review, Second Quarter
2003, 80(2), pp. 45-73.
Stock, James H. and Watson, Mark W. “Has Inflation
Become Harder to Forecast.” Prepared for the Board
of Governors of the Federal Reserve System conference Quantitative Evidence on Price Determination,
Washington, DC, September 29-30, 2005.
Stock, James H. and Watson, Mark W. “Forecasting
Output and Inflation: The Role of Asset Prices.”
Journal of Economic Literature, September 2003,
41(3), pp. 788-829.

McGuckin, Robert H.; Ozyildirim, Ataman and
Zarnowitz, Victor. “A More Timely and Useful
Index of Leading Indicators.” Journal of Business
and Economic Statistics, January 2007, 25(1),
pp. 110-20.

446

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

The Effectiveness of Monetary Policy
Robert H. Rasche and Marcela M. Williams
This analysis addresses changing views of the role and effectiveness of monetary policy, inflation
targeting as an “effective monetary policy,” monetary policy and short-run (output) stabilization,
and problems in implementing a short-run stabilization policy. (JEL E520)
Federal Reserve Bank of St. Louis Review, September/October 2007, 89(5), pp. 447-89.

T

he effectiveness of monetary policy
has been a long-standing question in
the monetary economics and central
banking literature. Perspectives on the
question have been influenced in part by developments in monetary theory and in part by interpretations of monetary history. Progress in the
discussion has also been influenced—indeed,
some might say hindered—by changing definitions of both “monetary policy” and “effectiveness.” Our discussion will address (i) changing
views of the role and effectiveness of monetary
policy, (ii) inflation targeting as an “effective
monetary policy,” (iii) monetary policy and shortrun (output) stabilization, and (iv) problems in
implementing a short-run stabilization policy.

CHANGING VIEWS ON THE
ROLE AND EFFECTIVENESS OF
MONETARY POLICY
What do analysts mean by “monetary policy”
and the “effectiveness” thereof? Each term is
something of a moving target. At times “monetary
policy” has referred to central bank actions to
influence and/or target some measure of the

money stock. Frequently, though certainly not
always, the definition of monetary policy has
focused on a measure of “high powered money”—
liabilities of the central bank. For a long time,
this was the definition incorporated in theoretical
models; in the policy arena this definition was the
foundation of the “monetarist revolution” in the
1960s and 1970s. A counter definition that was
likely the dominant perspective of policymakers
was that monetary policy referred to central bank
actions to influence and/or target short-term
interest rates or nominal exchange rates. Sargent
and Wallace (1975) advanced the proposition
that, in a model with “rational expectations,” the
price level (and all other nominal variables) could
be indeterminate if central banks set targets for
nominal interest rates, because the economy
would lack a “nominal anchor.” McCallum (1981)
showed that an appropriately defined interest rate
rule would avoid such indeterminacy. The interest
rate rule had to include a “nominal anchor.” In
recent years, in particular since Taylor’s (1993)
proposed characterization of Federal Open
Market Committee (FOMC) behavior in the early
Greenspan years, interest rate rules that include
a “nominal anchor” in the form of a desired or
target inflation rate have become the basic speci-

Robert H. Rasche is a senior vice president and director of research and Marcela M. Williams is a senior research associate at the Federal
Reserve Bank of St. Louis. This article was originally presented at the Bank of Korea’s conference, “The Effectiveness of Stabilization Policies,”
May 27, 2005, Seoul, Korea, and was published in their conference proceedings in 2006. The article is reprinted here with permission from
the Bank of Korea.

© 2007, The Federal Reserve Bank of St. Louis. Articles may be reprinted, reproduced, published, distributed, displayed, and transmitted in
their entirety if copyright notice, author name(s), and full citation are included. Abstracts, synopses, and other derivative works may be made
only with prior written permission of the Federal Reserve Bank of St. Louis.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

447

Rasche and Williams

fication of “monetary policy” in theoretical
analyses. (See, for example, Clarida, Galí, and
Gertler, 1999.)
The legacy of the Great Depression in the
United States and other industrialized economies
was that monetary policy was “ineffective.” This
perspective is most prominent in Keynes’s General
Theory and in the writings of the “Keynesian
economists” in the 1940s through the 1960s. For
example, the Radcliffe Committee in the United
Kingdom reported that
[t]he immediate object of monetary policy
action is to affect the level of total demand...
In theory, monetary action may work upon
total demand by changing the interest incentive; we believe that only very limited reliance
can be placed on this. More certainly, monetary
action works upon total demand by altering
the liquidity position of financial institutions
and of firms and people desiring to spend on
real resources; the supply of money itself is not
the critical factor. (Radcliffe, 1959, p. 135)

In the United States the minimalist perspective on the role and effectiveness of monetary
policy can be seen in the first two reports of the
Kennedy Council of Economic Advisers:
Unless the Government acts to make compensating changes in the monetary base, expansion
of general economic activity, accompanied by
increased demands for liquid balances and for
investment funds will tend to tighten interest
rates and restrict the availability of credit...
Discretionary policy is essential, sometimes to
reinforce, sometimes to mitigate or overcome,
the monetary consequences of short-run fluctuations of economic activity. In addition, discretionary policy must provide the base for
expanding liquidity and credit in line with
the growing potential of the economy.
(Council of Economic Advisers, 1962, p. 85)
Monetary policy as well as debt management
policy must be coordinated with fiscal policy
to secure the objectives of high employment
and growth without inflation. We are, and for
some time still will be, in a situation of substantial slack in labor force and capital resources,
a situation in which expansionary policies are
required...What matters most at this time is

448

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

2007

that financial policy should be designed to
facilitate rather than retard the expansionary
process which the tax program is designed to
launch. (Council of Economic Advisers, 1963,
p. 55)

A decade later, perspectives on the effectiveness of monetary policy had changed; and, in
some circles, monetary policy was viewed as
equally important as fiscal policy for affecting
both inflation and output fluctuations:
The past 10 years have been characterized by
an average growth rate of aggregate expenditures that is very high by historical standards
and that has substantially outstripped the sustainable growth of supply of real goods and
services. Contributing significantly to the
growth of aggregate demand were rapidly
increasing Government expenditures along
with monetary policies that were appreciably
more expansionary than those in earlier postWorld War II periods...When the inflationary
phase has lasted so long that expectations of
further inflation are firmly embedded in the
cost trend, a shift to policies of restraint first
exerts an adverse influence on output and the
desired price deceleration effect materializes
only with a lag. Any convincing interpretation
of the events during 1970 and 1973-4 must
stress this difficulty. (Council of Economic
Advisers, 1975, pp. 128-29)

This was not the only view of monetary and
fiscal policy at that time. The 1960s saw the rise
of “monetarism” subsequent to the work of
Friedman and Schwartz (1963), Friedman and
Meiselman (1963), and Andersen and Jordan
(1968). There were several planks in the monetarism platform. First and foremost was that sustained inflation was a monetary phenomenon and
that central banks should be held accountable for
maintaining price stability. Monetarists contended
that central banks should control the stock of
money in the economy, and not focus on targeting
short-term nominal interest rates, as the mechanism to achieve this long-run inflation objective.
The rationale for the focus on the growth of the
money stock was that, in a fiat money economy,
the money stock provided the nominal anchor
for the system.
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

In the eyes of monetarists, inflation control
was not the only concern of the monetary authorities. They saw monetary policy as having significant effects on short-run fluctuations in real output
(Andersen and Jordan, 1968; and Andersen and
Carlson, 1970), though not affecting long-run
output growth. Indeed, many monetarists (see
Brunner and Meltzer, 1968; Meltzer, 1976, 2003;
and Friedman and Schwartz, 1963) believed
monetary policy was responsible for aggravating,
not attenuating, historical cyclical fluctuations
in real output.
With the “rational expectations revolution”
in macroeconomics came the “policy ineffectiveness proposition” of the New Classical Macroeconomics (Sargent and Wallace, 1975). The
initial interpretations of this paradigm were that,
in any macroeconomic model, the assumption
of rational expectations would render monetary
policy ineffective in influencing real output, both
in the short run and long run. Hence, there was no
role for monetary policy in output stabilization.
Subsequent research (Fischer, 1977; Taylor, 1980;
and Calvo, 1983) demonstrated that it was the
interaction of the rational expectations hypothesis
and an assumption of perfectly flexible wages
and/or prices that generated the “policy ineffectiveness proposition.” The outgrowth of this
insight was the New Keynesian perspective.
With the widespread use of New Keynesian
models, the monetarist tenets about how “monetary policy” affects economic activity have become
widely held throughout academia and central
banking circles today, though most academics
and almost all central bankers would disown a
monetarist label. Money has largely disappeared
from discussions on monetary policy. Fry et al.
(2000, Chart 7.2, p. 123), using data from a Bank
of England survey, report that in the 1970s 11 (of
22) central banks in industrial countries reported
using a money and credit framework to formulate
monetary policy; but, by the 1990s, only 2 of these
banks maintained this framework. Von Hagen
(2004, Table 4) found a negative trend from 1970
to 2002 in the fraction of titles of articles in major
economics journals that included the word
“money,” though the frequency of titles including
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

“inflation” was relatively constant. He also found
that the frequency of “money” in the annual
reports of major central banks declined over the
period 1996-2002 (Von Hagen, 2004, Table 5).
King (2002) notes that
there is a paradox in the role of money in economic policy. It is this: that as price stability
has become recognized as the central objective
of central banks, the attention actually paid by
central banks to money has declined. (p. 162)
The decline and fall of money in policy formation is confirmed by a fall in the number of
references to money in speeches of central
bank governors. So much so that over the past
two years, Governor Eddie George has made
one reference to money in 29 speeches,
Chairman Greenspan one in 17, Governor
Hayami one in 11 and Wim Duisenberg three
in 30. (pp. 162-63)

In contemporary literature, models, and policy
discussions, attention is given to the role of an
inflation objective in a central bank “policy rule”
as the nominal anchor in a fiat money economy.
“Taylor rules” (Taylor, 1993) that specify a systematic relation between the target for a short-term
interest rate and deviations of inflation from an
inflation target and of real output from a measure
of “potential output” have become the norm for
the analysis of the impact of monetary policy.1
In this “rule like” environment, the setting of the
interest rate value is the policy action; the policy
itself is represented by the parameters of the
“rule,” including the inflation objective (π *) and
the respective weights that are assigned to deviations of observed inflation from that objective
and deviations of real output from “potential
output.”
1

There is an active discussion in the contemporary literature over
the design of monetary policy rules. Some economists (for example,
Svensson, 2005) argue for instrument rules that are optimized
from an objective function of the central bank and models of the
macroeconomy. Others (for example, McCallum and Nelson, 2004)
argue for independently constructed target rules in the spirit of
Taylor (1993). Both sides in this debate appear to accept the proposition that monetary policy can affect both inflation and real output in the short run, but only inflation in the long run. Hence, this
theoretical debate is about how to best implement monetary policy
rules, not about the effectiveness of policy.

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

2007

449

Rasche and Williams

Table 1
Inflation-Targeting Countries by Year of
Adoption
Year

Country

Total

1990

New Zealand
Chile

2

1991

Canada
Israel

2

1992

United Kingdom

1

1993

Sweden
Australia
Finland*

3

1994

Peru
Spain*

2

1995

—

0

1996

—

0

1997

—

0

1998

Czech Republic
Korea
Poland

3

1999

Mexico
Brazil
Colombia

3

2000

South Africa
Switzerland
Thailand

3

2001

Norway
Iceland
Hungary

3

2002

Philippines

1

NOTE: *Finland and Spain are considered to have become
non-inflation-targeting countries on joining the third stage of
the European Monetary Union in 1999.
SOURCE: Authors’ compilation based on monetary policy and
inflation reports of each country’s central bank and Mishkin
and Schmidt-Hebbel (2002) and Morande (2002).

LONG-RUN STABILIZATION
OBJECTIVES FOR MONETARY
POLICY
Over the past 15 years, a number of countries,
starting with New Zealand, have announced
explicit numeric inflation objectives (the π * term
in the Taylor rule framework). The relevant question is, How effective are central banks at hitting
explicit numeric inflation targets?
450

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

2007

Clearly, one straightforward way to address
this question is to look at the performance of
those countries that have announced explicit
numeric inflation targets. Our list of the countries
in this group (23 in all) is shown in Table 1. For
each country, Figure 1 graphs the year a target
was adopted and the inflation rate at that time.2,3
As far as we have been able to determine, no country has joined the group since 2002. In many, but
certainly not all of these countries, inflation was
below 5 percent at the time of the announcement
of the inflation-targeting regime. The appendix
provides the details on the target index(es), the
target ranges, dates of target adoption, target setting responsibility, and public reporting on the
performance of the inflation policy.
Figures 2 through 22 show the history of
inflation for each of the inflation-targeting countries.4 For those countries that have changed the
index in which they define their inflation target,
there is a separate graph for each index. The
graphs show the inflation target range (the shaded
area) or the point inflation target as appropriate.
It is immediately apparent from these graphs that
the period-to-period (month-to-month or quarterto-quarter) annualized rate of inflation is highly
volatile in all of the countries that pursue an
explicit numeric inflation target. These short-run
inflation rates are as likely as not to be outside the
target range. If effective monetary policy were to
be defined in terms of stability of high-frequency
rates of inflation, then all of these central banks
would have to be judged as failing to achieve the
objective. However, it is neither reasonable nor
desirable to define the objective in such short-run
terms. Shocks to the price level—that is, transitory shocks to inflation—originate from numerous
sources, both monetary and nonmonetary. No
central bank can foresee such shocks and probably
2

We consider two countries that at one time had explicit numeric
targets—Finland and Spain—that have dropped from the group
upon accession to the European Monetary Union.

3

This figure updates similar figures that can be found in Loayza
and Soto (2002) and Mishkin and Schmidt-Hebbel (2002).

4

For early discussions of the implementation of and experience
with inflation targeting for several of these countries, see
Leiderman and Svensson (1995).

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Figure 1
The Year of Adoption and Initial Inflation in Twenty-Three Inflation-Targeting Countries
Inflation (percent)
50
45

Peru

40
35
30

Chile

25
20

Israel

Hungary

Mexico

15
Czech Republic
10
5

Colombia
New Zealand

Poland
Canada

U.K.

Sweden
0
1988

1990

1991

Finland
Australia
1993

1994

cannot accurately predict the dynamics by which
such shocks ultimately impact the price level.
Economic theory suggests that central banks can
be held accountable for “sustained inflation.”
Correspondingly, inflation-targeting central banks,
and even central banks without explicit numeric
targets, such as the Fed, typically focus on
“medium term” inflation. The duration of the
medium term is frequently, and probably intentionally, left ambiguous.5 Absent a precise definition of the medium term, some measures must
be specified to judge the effectiveness of the
inflation-targeting policies.
We examine two measures of the effectiveness of explicit numeric inflation targeting. Both
measures are based on moving averages of the
observed rates of inflation. These measures are
5

President Santomero of the Federal Reserve Bank of Philadelphia
stated that he preferred measuring inflation against an explicit
target range on a 12-month moving-average (year-over-year) basis.
Few central bankers have been this explicit about their definitions
of a “medium term.”

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Korea

Spain

Brazil
South Africa

1995

1997

Philippines
Iceland
Norway
Thailand

Switzerland

1998

1999

2001

2002

indicated in Figures 2 to 23 by the heavy lines.
Relative to the dates indicated on the horizontal
axis, this line shows the leading moving average
of the rate of inflation to the end of 2004. The
shortest moving average shown is one year. The
question is, What is the maximum period, ending
with 2004, that the moving average of the inflation rate remained within the bounds determined
by the current (end of 2004) inflation target?
These periods are shown for each of the inflationtargeting countries in Table 2.
Judged by this metric, there are a number of
inflation-targeting countries in which monetary
policy has been very effective. For five countries,
New Zealand, Norway, Switzerland, Thailand,
and the United Kingdom, the moving average of
the rate of inflation has been within the current
announced target range since before the adoption
of the inflation-targeting procedure.6 Canada and
6

Norway has a stated target of 2.5 percent, not an inflation range.
The moving average of the inflation rate has been below this value
since 2000 and only dipped slightly below zero in early 2003.

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

2007

451

Rasche and Williams

Figure 2
Australia: Moving Average of Inflation and Inflation Target Range
A. Treasury Underlying Consumer Price Index (CPI)
Inflation (percent)
7

Inflation Target Range
Quarterly Treasury Underlying Inflation

6

Leading Moving Average of Inflation

5
4
3
2
1
0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

B. Headline CPI
Inflation (percent)
16

Inflation Target Range

14

Quarterly Headline CPI
Leading Moving Average of Inflation

12
10
8
6
4
2
0
1990
–2

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–4

NOTE: The initial, formal inflation target related to the Treasury underlying measure of the CPI. When interest charges were removed
from the CPI in the June quarter 1998, this target was amended to the headline CPI inflation rate.
SOURCE: Reserve Bank of Australia.

452

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Figure 3
Brazil: Moving Average of Inflation and Inflation Target Range
Inflation (percent)
Inflation (percent)

90

40
35

80

30
25

70

20

60

15
10

50

5
0
2000
–5

40

2001

2002

2003

2004

30
20

Inflation Target Range
Monthly Total CPI

10

Leading Moving Average of Inflation
0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–10
–20

SOURCE: Banco Central do Brazil.

Figure 4
Canada: Moving Average of Inflation and Inflation Target Range
Inflation (percent)
35

Inflation Target Range
Monthly Total CPI

30

Leading Moving Average of Inflation
25
20
15
10
5
0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–5
–10
–15

SOURCE: Banque du Canada.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

453

Rasche and Williams

Figure 5
Chile: Moving Average of Inflation and Inflation Target Range
Inflation (percent)
70
Inflation Target Range

Inflation (percent)
15

Monthly Total CPI Inflation

60

10

Target Midpoint
50

Leading Moving Average of Inflation

5

40

0
2000

30

2001

2003

2002

2004

–5

20

–10

10
0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–10
–20

SOURCE: Banco Central de Chile.

Figure 6
Colombia: Moving Average of Inflation and Inflation Target Range
Inflation (percent)

Inflation Target Range

60

Monthly Total CPI Inflation

30

Target Midpoint

25

Leading Moving Average of Inflation

20

50

Inflation (percent)

15
10

40

5
0
2000

30

2001

2002

2003

2004

–5

20

10

0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–10

SOURCE: Banco Central de Colombia.

454

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Figure 7
Czech Republic: Moving Average of Inflation and Inflation Target Range, Headline CPI
Inflation (percent)
60

50

Inflation (percent)

Inflation Target Range

25

Monthly Total CPI

20
15

Leading Moving Average of Inflation

10
5

40

0
2000

2001

2002

2003

2004

–5

30

–10
–15

20

10

0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–10

–20

NOTE: Czech Republic targeted net inflation through 2001 and headline CPI thereafter.
SOURCE: Czech Statistical Office.

Figure 8
Hungary: Moving Average of Inflation and Inflation Target Range
Inflation (percent)
90

Inflation (percent)
30
25

80

20
15

70

10

60

5

50

–5

0
2000

2001

–10

2002

2003

2004

2005

Inflation Target Range

40

Monthly Total CPI
30

Leading Moving Average of Inflation

20
10
0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–10

SOURCE: Hungarian Central Statistical Office.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

455

Rasche and Williams

Figure 9
Iceland: Moving Average of Inflation and Inflation Target Range
Inflation (percent)
25

Inflation Target Range
Monthly Total CPI
Leading Moving Average of Inflation

20

15

10

5

0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

–5

–10

SOURCE: Statistics Iceland.

Figure 10
Israel: Moving Average of Inflation and Inflation Target Range
Inflation (percent)
40

Inflation Target Range
Monthly Total CPI Inflation
Target Midpoint

30

Leading Moving Average of Inflation
20

10

0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–10

–20

SOURCE: Central Bureau of Statistics, Israel.

456

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Figure 11
Korea: Moving Average of Inflation and Inflation Target Range
A. Total CPI Inflation
Inflation (percent)
35

Inflation Target Range
Monthly Total CPI Inflation

30

Target Midpoint
Leading Moving Average of Inflation

25
20
15
10
5
0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–5
–10

B. Core CPI Inflation
Inflation (percent)
30

Inflation Target Range
Monthly Core CPI Inflation

25

Target Midpoint
Leading Moving Average of Inflation

20

15

10

5

0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–5

NOTE: Total CPI was targeted until 1999. Core inflation was targeted beginning in 2000.
SOURCE: Korea National Statistics Office and the Bank of Korea.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

457

Rasche and Williams

Figure 12
Mexico: Moving Average of Inflation and Inflation Target Range
Inflation (percent)

Inflation (percent)

120

20

15

100
10

80

5

0
2000

2001

2002

2003

2004

60
–5

40

20

0
1990
–20

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

Inflation Target Range
Monthly Total CPI Inflation
Target Midpoint
Leading Moving Average of Inflation

SOURCE: Banco de Mexico.

458

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Figure 13
New Zealand: Moving Average of Inflation and Inflation Target Range
A. CPI, Excluding Credit Services (CPIX)
Inflation (percent)
8

Inflation Target Range
Monthly CPI, Excluding Credit Services

7

Leading Moving Average of Inflation

6
5
4
3
2
1
0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–1

B. Headline CPI
Inflation (percent)
8

Inflation Target Range
Monthly Total CPI
Leading Moving Average of Inflation

6

4

2

0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–2

–4

NOTE: The Reserve Bank of New Zealand targeted underlying inflation until the September 1997 quarter, CPIX inflation through to
the June 1999 quarter, then CPI inflation from the September 1999 quarter.
SOURCE: Reserve Bank of New Zealand.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

459

Rasche and Williams

Figure 14
Norway: Moving Average of Inflation and Inflation Target Range
Inflation (percent)
35

Total CPI

30

Inflation Target
Leading Moving Average of Inflation

25
20
15
10
5
0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–5
–10
–15
–20

SOURCE: Statistics Norway.

Figure 15
Peru: Moving Average of Inflation and Inflation Target Range
Inflation (percent)
100

Inflation (percent)
15

10

80
5

60
0
2000

2001

2002

2003

2004

–5

40

Inflation Target Range
–10

Monthly Total CPI

20

0
1990

Leading Moving Average of Inflation

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–20

SOURCE: Central Reserve Bank of Peru.

460

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Figure 16
Philippines: Moving Average of Inflation and Inflation Target Range
Inflation (percent)

Inflation (percent)

90

25

20

70

15

10

50
5

30

0
2000

2001

2002

2003

2004

2005

–5

10
1990
–10

1991 1992

1993 1994

1995 1996

1997 1998

1999 2000

2001 2002

2003 2004

Inflation Target Range
Monthly Total CPI Inflation

–30

Target Midpoint
Leading Moving Average of Inflation

–50

SOURCE: Central Bank of the Philippines.

Figure 17
Poland: Moving Average of Inflation and Inflation Target Range
Inflation (percent)
180

Inflation (percent)

Inflation Target Range

25

Monthly Total CPI Inflation
160

20

140

15

120

10

100

5

Target Midpoint
Leading Moving Average of Inflation

80

0
2000

60

–5

40

–10

2002

2001

2003

2004

20
0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–20

SOURCE: Central Statistical Office, Poland.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

461

Rasche and Williams

Figure 18
South Africa: Moving Average of Inflation and Inflation Target Range
Inflation (percent)
25

Inflation Target Range
Monthly CPIX Inflation
Leading Moving Average of Inflation

20

15

10

5

0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–5

–10

SOURCE: South African Reserve Bank.

Figure 19
Sweden: Moving Average of Inflation and Inflation Target Range
Inflation (percent)
50

Inflation (percent)
15

Inflation Target Range
Monthly Total CPI Inflation

40

10

Leading Moving Average of Inflation
5

30

0
2000

2001

2002

2003

2004

–5

20

–10

10
–15

0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–10

–20

SOURCE: Statistics Sweden.

462

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Figure 20
Switzerland: Moving Average of Inflation and Inflation Target Range
Inflation (percent)
20

Inflation Target Range
Monthly Total CPI Inflation

15

Leading Moving Average of Inflation

10

5

0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–5

–10

–15

SOURCE: Swiss National Bank.

Figure 21
Thailand: Moving Average of Inflation and Inflation Target Range
Inflation (percent)
20

Inflation Target Range
Monthly Core CPI Inflation
Leading Moving Average of Inflation

15

10

5

0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–5

–10

SOURCE: Bureau of Trade and Economic Indices, Ministry of Commerce, Thailand.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

463

Rasche and Williams

Figure 22
United Kingdom: Moving Average of Inflation and Inflation Target Range
A. Retail Price Index, Excluding Mortgage-Interest Payments (RPIX)
Inflation (percent)
50

Inflation (percent)
12
10
8

RPIX Inflation Target Range

40

6

Monthly RPIX Inflation

4

RPIX Lower Bound

30

2
0
2000
–2

Leading Moving Average of Inflation
20

2001

2002

2003

2004

–4
–6
–8

10

0
1990

–10

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–10

–20

B. CPI
Inflation (percent)

Inflation (percent)

50

15
10

40

5
0
2000

30

2001

2002

2003

2004

–5
–10

20

–15

10

0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

–10
CPI Inflation Target Range
–20

Monthly Total CPI Inflation
Total CPI Inflation
Leading Moving Average of Inflation

NOTE: Since December 2003, the United Kingdom’s inflation target has been based on the CPI.
SOURCE: National Statistics, United Kingdom.

464

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Figure 23
United States: Moving Average of Inflation and Inflation Target Range
A. Core CPI
Inflation (percent)
9

Inflation Target Range
Monthly Core CPI Inflation

8

Leading Moving Average of Inflation
7
6
5
4
3
2
1
0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

B. Core PCE
Inflation (percent)
12

1-3 Percent Inflation Range
Monthly Core PCE Inflation

10

Leading Moving Average of Inflation
8
6
4
2
0
1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

–2
–4
–6
–8

NOTE: CPI for all urban consumers: All items less food and energy.
SOURCE: FRED®, Federal Reserve Bank of St. Louis.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

465

Rasche and Williams

Table 2
Effectiveness of Explicit Numeric Inflation Targeting

Country
Australia
Brazil
Canada

First date for
Target range as of
which target
Dec. 2004 (%)
range is not met

Number of
periods
within range

Moving average
rate of inflation
from Dec. 2002
to Dec. 2004 (%)

Standardized
deviation
from
moving average

17

2.51

0.02

2-3

June 2000

4.5 ± 2.5

Dec. 2003

12

8.80

1.72

2±1

Nov. 1992

109

2.24

0.24

Chile

2-4

Nov. 2003

13

1.76

–1.24

Colombia

5-6

Dec. 2002

24

6.17

1.34

1-3

Feb. 2002

34

1.93

–0.07

Hungary

Czech Republic

3.5 ± 1

Dec. 2003

12

5.75

2.25

Iceland

2.5 ± 1.5

Aug. 2003
(next date is April 2001)

16

3.37

0.58

Israel

1-3

Nov. 2003

13

–0.35

–2.35

Korea

2.5-3.5

Nov. 2003
(next date is Nov. 2001)

13

2.87

–0.26

Mexico

2-4

Dec. 2003

12

4.69

1.69

New Zealand

1-3

Never 1

59

2.15

0.15

2.5

Never

45

0.85

—

2.5 ± 1

Nov. 2003

13

3.03

0.53

Norway 2
Peru
Philippines
Poland
South Africa
Sweden

4-5

Dec. 2003

12

6.38

3.76

2.5 ± 1.5

Dec. 2003

12

3.09

0.39

3-6

Feb. 2002

34

4.25

–0.17

2±1

Dec. 2003

12

0.78

–1.22

<2

Never 3

59

0.97

–0.03

0-3.5

Never 4

55

0.30

–0.83

2±1

Never 5

146

1.46 6

–0.54

Switzerland
Thailand
United Kingdom

NOTE: 1 Average rate of inflation for New Zealand has never gone outside the range since inflation targeting was adopted (1990).
2 No index numbers for CPI-ATE (CPI, adjusted for tax changes and excluding energy products) are available. Average rate of headline
CPI inflation was used instead. Value for average rate of headline CPI inflation has never exceeded 2.5% since inflation targeting was
adopted (2001). 3 Average rate of inflation for Switzerland has never gone outside the range since inflation targeting was adopted (2000).
First date for which the average rate of inflation was outside the range was October 1989 at 2.09%. 4Average rate of inflation for Thailand
has never gone outside the range since inflation targeting was adopted (2000). The first date for which the average rate of inflation
was outside the range was August 1992 at 3.52%. 5 First date for which the average rate of inflation was outside the range was March
1992 at 3.03% for average rate of RPIX inflation. 6 Note that in December 2002, the United Kingdom targeted RPIX and not headline
CPI. The value for the average rate of inflation of RPIX at that time was 2.6%, still within the target range as of December 2004.

466

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Australia also have maintained, for considerable
periods, an average rate of inflation within the
range currently in effect. (Note that for Australia
and New Zealand, data are quarterly and not
monthly.) None of these countries has a particularly wide target range. The Czech Republic and
South Africa have shorter records of success by
this metric, but nevertheless have moderately
effective performances. The Czech Republic is
notable because the inflation rate there was fairly
high when the target was adopted and the moving
average of the rate of inflation has fallen outside
of the target range only on the low side. Nevertheless, the moving average of Czech headline
inflation has been positive for the entire period
since it fell below the lower bound of the current
target range (in February 2002).
Israel, Peru, and Poland have experienced
long-term average inflation below their current
target ranges. The short-horizon moving averages
for Hungary, Peru, and Poland have exceeded
the current target ranges. In Israel, the moving
average of the rate of inflation actually went negative in 2002 and 2003. Three countries—Chile,
Colombia, and Hungary—have adjusted their
targets downward over time, and generally the
average inflation rate has fallen below the ranges,
consistent with success in moving to the lower
inflation targets. Deviations from the target ranges
have been symmetric. Other countries, notably
Brazil, Mexico, and the Philippines, have consistently missed their target ranges on the high side.
Although averages provide interesting insights
into the sustainability of inflation performance,
they obscure the marginal performance. A moving
average could remain within the target range for
a long period of time if, over time, the inflation
rate converges toward the midpoint of the range.
Alternatively, the same moving average could
result if, early in the period, the inflation rate was
close to one end point of the target range and, as
time progressed, inflation moved close to the
opposite edge of the target range. The latter situation could be characterized as “skating on thin
ice.”
To examine this issue, Table 2 shows the value
of the moving average over the two-year period
2003-04 and a standardized deviation of this twoF E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

year moving average from the midpoint of the
target range that prevailed at the end of 2004.
The standardization is constructed by dividing
the deviation of the moving average from the
midpoint by one-half the difference between the
upper and lower endpoints of the target range.
By this metric, the bulk of the inflationtargeting countries have been doing quite well
over the past two years. The exceptions are Brazil,
Colombia, Hungary, Mexico, and the Philippines
(whose average inflation rate over the past two
years fell above their target ranges) and Israel and
Sweden (whose average inflation rate over the
past two years fell below the target ranges).
Our conclusion from these data is that central
banks that have announced explicit numeric
inflation objectives have been quite effective in
achieving the stated inflation stabilization
objective.
The FOMC has not adopted this framework,
though it is known that on at least three occasions
the pros and cons of adopting this approach have
been debated around the FOMC table.7 Several
current participants in the FOMC have stated on
the record their preference for an explicit numeric
target and given their preferred measures. Included
are then-Governor Ben Bernanke, President
Jeffrey Lacker of the Federal Reserve Bank of
Richmond, President Janet Yellen of the Federal
Reserve Bank of San Francisco, and formerPresident Anthony Santomero of the Federal
Reserve Bank of Philadelphia.8
Then-Governor Bernanke indicated his preferred inflation target is 1 to 2 percent as measured
by the core personal consumption price index.9
President Lacker has indicated his preference for
7

Transcripts of two of these debates, on January 31, 1995, and July 2,
1996, can be found on the web site of the Board of Governors of
the Federal Reserve System: www.federalreserve.gov/fomc/
transcripts/transcripts_1995.htm and www.federalreserve.gov/
fomc/transcripts/transcripts_1996.htm, respectively. A summary
of the most recent debate at the February 1, 2005, FOMC meeting
is also available on the Board’s web site: www.federalreserve.gov/
fomc/minutes/20050202.htm.

8

During the July 1996 FOMC debate on inflation targets, President
Gary Stern of the Federal Reserve Bank of Minneapolis indicated
that a 2 percent target (in terms of the CPI) would be acceptable to
him (FOMC Transcripts, July 2-3, 1996, p. 56). The discussion at
that time was not framed in terms of a target point or a target range.

9

See, for instance, Derby (2005).

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

2007

467

Rasche and Williams

a target of 2 percent as measured by the core CPI
or 1.5 as measured by the core personal consumption price index; he has also indicated a preference
that inflation be kept above 1 percent (Lacker,
2005). President Yellen has indicated a preference
for a target of 1.5 percent as measured by the core
personal consumption price index, with a range
of about ± 1 percent (Reuters News, 2005). FormerPresident Santomero indicated his preference
for a target range of 1 to 3 percent as measured
by a 12-month moving-average rate of change in
the core personal consumption price index (AFX
Asia, 2004). Other current participants at FOMC
meetings, including Vice Chairman Donald Kohn
(2005), have indicated that they do not prefer an
explicit numeric inflation objective. President
William Poole has stated that he believes “ambiguity with respect to the Fed’s inflation and
employment objectives is not large and is not the
main problem the Fed faces with its communication policies” (Poole, 2005b).
Differences of opinion among FOMC participants notwithstanding, in May 2003 the press
release following the FOMC meeting indicated
that “the probability of an unwelcome substantial
fall in inflation, though minor, exceeds that of a
pickup in inflation from its already low level.10
The minutes of that FOMC meeting (FOMC
Minutes, May 6, 2003) indicate the rationale for
this statement:
Members commented that substantial additional disinflation would be unwelcome
because of the likely negative effects on economic activity and the functioning of financial
institutions and markets, and the increased
difficulty of conducting an effective monetary
policy, at least potentially in the event the
economy was subjected to adverse shocks.
Members also agreed that there was only a
remote possibility that the process of disinflation would cumulate to the point of a decline
for an extended period in the general price
level.11

At that time, core personal consumption
inflation was measured in the neighborhood of
10

See www.federalreserve.gov/boarddocs/press/monetary/2003/
20030506/default.htm.

11

See www.federalreserve.gov/fomc/minutes/20030506/htm.

468

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

2007

1 percent. Although the FOMC has never stated
a numeric inflation objective, individual FOMC
participants have expressed preferences for both
the core CPI and core personal consumption price
index; they have typically indicated values for
core CPI inflation one-half percent above those
for the core personal consumption inflation rate.
It seems reasonable to conclude that the FOMC
has a lower bound of an acceptable medium-term
rate of inflation in the neighborhood of 1 percent
for the core personal consumption inflation rate
and perhaps 1.5 percent for the core CPI inflation
rate.
Former-Governor Larry Meyer (2004) is also
on record in favor of an explicit numeric inflation
objective. However, his position is that the “dual
mandate” inherent in the Federal Reserve Act
differentiates the U.S. environment from that of
other inflation-targeting central banks that operate under a “hierarchical mandate.” Meyer defines
a hierarchical mandate as an environment “in
which price stability is identified as the principal
objective, and central banks are restricted from
pursuing other objectives unless price stability
has been achieved” (p. 151). He contrasts this
with the “dual mandate,” where “monetary policy
is directed at promoting both full employment
and price stability with no priority expressed,
and with the central bank responsible for balancing these objectives in the short run” (p. 151). It
is our opinion that Meyer’s view does not allow
for the effectiveness of monetary policy to vary
in the long and short runs. In terms of long-run
objectives, central banks must necessarily operate under a hierarchical mandate, given the consensus view of monetary policy that policymakers
are not presented a long-run tradeoff between
inflation and real output. Indeed, in specifying a
policy rule, whether an instrument or target rule,
the exercise of determining how much weight to
place on short-run movements in inflation versus
short-run movements in real output is conditioned
on the prespecification of the long-run inflation
target (π *). In this sense, any central bank seeking
to operate in such a monetary policy framework
has to be hierarchical: First it must specify its
long-run inflation objective and then, and only
then, can it set its preferred (or optimal) weights
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

for short-run fluctuations.12 The choice of weights
could be such that the central bank follows a
hierarchical mandate in both the long and short
runs; however, there is nothing to preclude pursuing a dual goal, with a short-run mandate
nested within a hierarchical long-run mandate.
It is likely that most, if not all, central banks that
have adopted an explicit inflation target pursue
that objective within a nested hierarchical/dual
structure.
Panels A and B of Figure 23 show the core CPI
inflation and core personal consumption price
inflation for the United States; the leading moving average from each of the dates since January
1990 until the end of 2004 are also shown.13 The
shaded area, from 1 to 3 percent in core CPI inflation, appears to encompass the preferences of
the FOMC participants who have spoken out in
favor of an explicit numeric inflation objective.
The leading moving-average rate of core CPI inflation in the United States bottomed out in August
2002 at a value of 1.64 percent (annual rate). The
corresponding date and value for core personal
consumption price inflation are December 2002
and 1.28 percent, which appear to be close to the
bottom of the FOMC’s implicit acceptable range
of inflation. On the other end of the scale, the
leading moving-average rate of core CPI inflation
has been below the 3 percent level since March
1991, whereas that for personal consumption price
inflation has been below the 3 percent level since
March 1987. These are comparable to the best
performance of the inflation-targeting central
banks against their announced targets. Accordingly, it cannot be claimed that an explicit numeric
inflation target is a necessary condition to produce
low and stable rates of inflation for an extended
period. The question, which will not be answered
unless inflation pressures build in the future, is
whether in the absence of a public numeric inflation objective the institutional commitment exists
to take potentially unpopular policy actions to
resist upward creep in inflation.
12

See also Svensson (2004).

13

Relative to the end of 2004, the line indicates a trailing moving
average of inflation back to the date indicated.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

HOW EFFECTIVE ARE CENTRAL
BANKS AT SHORT-RUN (OUTPUT)
STABILIZATION?
The evidence on the effectiveness of monetary
policy as a short-run stabilization device is problematic. As Poole (2005b) has noted,
[t]he only certainty is that the effect of policy
actions on real variables eventually dissipates.
“Eventually” may cover a period of several
years, and may be longer in some circumstances than others. It is worth noting that these
hedges on my part reflect ignorance—mine and
the profession’s—and not obfuscations. We just
don’t have precise estimates of the magnitudes
and durations of effects of monetary policy on
real variables.

Our objective here is to examine why a definitive answer to this question remains so illusive.
On one hand there is “case study” evidence supporting the idea that monetary policy does affect
output fluctuations in the short run. The most
prominent evidence from such studies highlights
the contractionary effects of monetary policy. On
the other hand, there are volumes of VAR analyses
that fail to determine a major role for monetary
policy in short-run stabilization.
The best known, though not uncontested,
case-study analysis of the short-run response of
real activity to monetary policy is Friedman and
Schwartz’s (1963) monetary history. They argue
that the Federal Reserve put the “great” in the
Great Contraction:
The monetary character of the contraction
changed drastically in late 1930, when several
large bank failures led to the first of what were
to prove a series of liquidity crises involving
runs on banks and bank failures on a scale
unprecedented in our history...
The drastic decline in the stock of money and
the occurrence of a banking panic of unprecedented severity did not reflect the absence of
power on the part of the Reserve System to prevent them. Throughout the contraction, the
System had ample powers to cut short the
tragic process of monetary deflation and banking collapse. Had it used those powers effectively in late 1930 or even in early or mid-1931,

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

2007

469

Rasche and Williams

the successive liquidity crises that in retrospect
are the distinctive feature of the contraction
could almost certainly have been prevented
and the stock of money kept from declining,
or indeed, increased to any desired extent.
Such action would have eased the severity of
the contraction and very likely would have
brought it to an end at a much earlier date.
(pp. 10-11)

Romer and Romer (1989) construct case
studies of six episodes from World War II through
1979 in which they believe that the Fed deliberately took action to induce a recession to reduce
inflation. They conclude that the evidence supports the hypothesis that the monetary policy
actions had a significant negative effect on real
output in all of these instances. Case studies such
as these address the qualitative question of
whether monetary policy has an effect on real
output; they do not address the question of the
magnitude of the output response to a change in
policy.
The final experience that is widely cited as
evidence of a contractionary effect of monetary
policy is the U.S. experience in 1979-83: the socalled “Volcker disinflation.” This period is
marked by two separate recessions: January–July
1980 and July 1981–November 1982. The first
recession followed closely the introduction of
the “new operating procedures” in October 1979
and an increase of 6 percent in the federal funds
rate.14 Note that the increase in the funds rate
was not directly targeted by the Fed under the
“new operating procedures.” Furthermore, the
impact of the monetary policy action in 1980 is
confounded with the introduction of credit controls by the Carter administration in March 1980.15
Goodfriend (2005) maintains that the recession
of 1981-82 was the direct consequence of monetary policy directed at disinflation:
The lesson of 1980 was that the Fed could not
restore credibility for low inflation if it continued to utilize interest rate policy to stabilize
14

15

For an analysis of the environment that led to the introduction of
the “new operating procedures” and the objectives that the
Volcker Fed sought to achieve with this innovation, see Lindsey,
Orphanides, and Rasche (2005).
See Schreft (1990).

470

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

2007

the output gap...As measured by personal
consumption expenditures (PCE) inflation,
which was about 10 percent in Q1 1981, real
short-term interest rates were then a very high
9 percent. Not surprisingly, the aggressive
policy tightening began to take hold by midyear. (pp. 316-17)

Certainly, the home building industry in the
United States regarded the collapse of housing
construction during both recessions as the direct
responsibility of the Volcker Fed, as evidenced
by the numerous complaints delivered to the
Board of Governors written on 2 × 4s. The housing
construction industry in the United States showed
highly cyclical fluctuations through the recession
of 1990-91 (see Figure 24), and concerns about
the sensitivity of this industry to monetary policy
actions had been the focus of discussion at least
since the early 1960s.16
Housing starts and housing construction
behaved very differently in the 2001 recession
than in prior postwar recessions: No slowdown is
obvious. Admittedly, this cyclical slowdown was
very mild, at least as measured in terms of real
output growth. Yet this raises the question of
whether cyclical fluctuations in housing should
be cited as universal evidence of an impact of
monetary policy on short-run fluctuations.17
One legacy of the Great Depression in the
United States has been the use of price controls
on bank deposits—so-called Reg Q ceilings. In
1966 these controls were extended to liabilities
of thrift institutions that, at the time, were the
principal source of mortgage financing. Cyclical
fluctuations in interest rates had a major impact
on the availability of mortgage financing during
this period. By the mid-1980s these price controls
had been removed, but by that time (economic)
insolvency was widespread among thrift institutions. The resolution of the crisis in the housing
finance industry continued through the recession
of 1990-91. Hence, it may be more appropriate to
argue that the interaction of monetary policy with
16

See, for example, Federal Reserve Bank of Boston (1970) and
Grebler and Maisel (1963).

17

Stock and Watson (2003) note the large decline in the volatility of
residential construction (though not nonresidential construction)
in the United States since the mid-1980s (p. 39).

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Figure 24
U.S. Housing Starts: Annual Rates
Thousands of Units
3,000

2,500

2,000

1,500

1,000

500

0
1959

1962

1965

1968

1971

1974

1977

1980

1983

1986

1989

1992

1995

1998

2001

2004

NOTE: Shaded bars indicate recessions.

the system of deposit price controls produced a
unique environment that supported a cyclical
response of the economy to monetary policy
actions. In the current U.S. environment, where
mortgage securitization has become the rule and
specialized deposit intermediaries have ceased
to be significant players in mortgage finance, a
traditional argument for the transmission of monetary policy may be more tenuous.
Econometric models provide alternative evidence on the effectiveness of monetary policy in
influencing the short-run behavior of real output.
Over the past 25 years, since the publication of
Sims’s (1980) classic article, literally hundreds,
perhaps thousands, of econometric studies in
vector autocorrelation (VAR) frameworks have
sought to address this question. We believe that
few people would argue that research in this
framework has provided conclusive evidence to
support the hypothesis that monetary policy has
strong short-run effects on real output fluctuations.
Christiano, Eichenbaum, and Evans (1999) summarize their extensive overview of this literature:
F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

“[V]iewed across both sets of identification strategies that we have discussed, there is a great deal
of uncertainty about the importance of monetary
policy shocks in aggregate fluctuations” (p. 127)
and “there is agreement that monetary policy
shocks account for only a very modest percentage
of the volatility of aggregate output; they account
for even less of the movements in the aggregate
price level”(p. 71). But if a consensus from case
studies of historical episodes is that there are substantial effects, the question is how to reconcile
the apparently conflicting evidence. An early
assessment of the VAR type of study is provided
by Cagan (1989):
If we accept the bulk of historical evidence as
confirming the important monetary effects on
the real economy, contrary findings cannot be
fully valid. And, if such contrary evidence is
not valid, what kind of evidence in monetary
research is acceptable and convincing? (p. 119)
The VAR seems to me to be hopelessly unreliable and low in power to detect monetary
effects of the kind we are looking for and

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

2007

471

Rasche and Williams

believe, from other kinds of evidence, to exist.
(p. 127)

In the approximately 15 years since Cagan
posed this question, analysts have become much
more aware of the limitations of VAR analyses. It
is now well understood that the VAR approach
does not solve the fundamental econometric problem of identification. The VARs that are readily
estimated using standard econometric software
are no more than reduced-form models.18 Indeed,
there is substantial risk of misspecification as a
result of omitted variables, given the limits on
the dimensionality of the typical VAR that is
imposed by the available time span of macroeconomic data series.
In the formative years of VAR analysis (say
1980-86) the typical approach was to “rotate and
orthogonalize shocks” by computing a Cholesky
decomposition of the covariance matrix of the
estimated VAR residuals and to assume that one of
the resulting “shocks,” frequently that associated
with a short-term interest rate, represented the
monetary policy innovation—the unpredictable
component of monetary policy. Analyses of the
effectiveness of monetary policy were constructed
from impulse response functions and variance
decompositions with respect to this “monetary
shock.”
Gradually, it became recognized that “recursiveness and orthogonalization” is the imposition
of a particular set of identifying restrictions—a
triangular Wold causal chain structure.19 This
approach to identification was widely rejected
by the econometrics establishment when initially
proposed in the 1960s. Starting in the mid-1980s,
alternative restrictions for identification of
“structural VARs” (SVARs) appeared in the literature.20 Generally the SVAR framework has maintained the identifying restrictions that the shocks
in the “economic model” are independent and
has found the additional required restrictions
among the only available alternatives: constraints
on impact or steady-state multipliers of the SVAR
18

For an extensive discussion of the identification problem in VAR
models, see Christiano, Eichenbaum, and Evans (1999, Section 2).

19

See Wold (1954 and 1960).

20

See, for example, Sims (1986) and Bernanke (1986).

472

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

or exclusion restrictions on the slope coefficients
among contemporaneous variables or steadystate relationships in the SVAR.
Steady-state identifying restrictions are those
for which accepted theory provides the most
insight. Such restrictions may provide information on the dynamics of a real output response to
a monetary policy shock that produces a permanent change in the inflation rate (assuming that
inflation is approximately a nonstationary variable
during the sample period). This is facilitated by
received macroeconomic theories that suggest
only monetary shocks can produce sustained
changes in inflation. In an economy where the
central bank focuses on a rule for an interest rate
target that responds to deviations from a desired
rate of inflation and other variables such as output
gaps, such monetary shocks occur only when
there is a change in the inflation target.21 This
does not get to the question of the effectiveness
of monetary policy for short-run output stabilization. Here, the issue is how real output responds
to monetary shocks that cause transitory fluctuations in the inflation rate (i.e., changes in the
price level).
Unfortunately, received macroeconomic theories suggest that shocks from many nonmonetary
sources can have a permanent effect on the price
level. Examples include fiscal policy shocks,
energy price shocks, productivity shocks, and
terms-of-trade shocks. In such economic structures, restrictions on impact multipliers are hard
to justify and sufficient restrictions on slope
coefficients among the contemporaneous variables in the VAR to identify the desired monetary
shock are problematic. This concern is echoed in
Romer and Romer (1989, p. 121):
The reason that purely statistical tests, such
as regressions of output on money, studies of
the effects of “anticipated” and “unanticipated”
money, and vector autoregressions, probably
have not played a crucial role in forming most
economists’ views about the real effects of
monetary disturbances is that such procedures
cannot persuasively identify the direction of
causation.
21

2007

This conclusion should hold regardless of whether the central
bank pursues an instrument rule or a target rule.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Identification of the effectiveness of monetary
policy to stabilize output fluctuation is further
complicated by a lack of transparency and likely
a lack of stationarity in the rule-like behavior of
central banks. There is an ongoing debate about
whether FOMC behavior over a long period can
be characterized by a common rule-like specification. Romer and Romer (2002a,b) argue that
the actions of the FOMC in the 1950s and in the
1980-90s were similar in their rule-like characteristics, but that during the 1960s and 70s a different “regime” was in place. Orphanides (2001
and 2002) and Orphanides and van Norden (2002)
argue that, when judged in terms of real-time data,
the rule-like behavior of the FOMC in the 1960s
and 1970s is consistent with behavior in the 1980s
and 1990s. They conclude that the Great Inflation
did not result from bad policy, but from applying
reasonable policy without recognition of and
adjustment for biased measurements of “potential
output.” Either view of the 1960 and 1970s poses
a challenge to the standard approach of identifying monetary shocks in SVAR structures.
Beyond the arguments about the specification
of monetary policy during the Great Inflation,
there are other concerns that at least occasionally
dominate central bankers’ discussions of policy
objectives. Such concerns at a minimum contaminate efforts to identify policy rules with measurement error and likely also contaminate the
assumed identifying restrictions.
For the FOMC, there are at least four incidents
in the past 20 years (documented in the published
record of FOMC minutes and transcripts) in which
concerns about financial stability dominated
policy decisions and policy actions were driven
by issues in addition to inflation or output stabilization. These incidents include the stock market
collapse in October 1987, the Asian crisis/Russian
default in August-October 1998, Y2K in late 1999,
and the 9/11 tragedy in September 2001. Some
analysts add the credit crunch/financial headwinds concern in 1990-93 to this list.22
According to the unofficial staff interpretations
of FOMC policy changes compiled by Thornton
and Wheelock (2000), the expected funds rate was
22

See, for example, Romer and Romer (2002b, p. 68).

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

decreased by 37.5 basis points on October 23,
1987, and by an additional 12.5 to 25 basis points
on October 28, 1987, in response to the stock
market crash. This interrupted the succession
of increases in the expected funds rate that had
started on January 15, 1987. Increases in the
expected funds rate were not resumed until
March 29, 1988, roughly six months after the
crash. During a conference call on October 20,
1987, Chairman Greenspan noted,
I think we’re playing it on a day-to-day basis.
And in a crisis environment. I suspect we
shouldn’t really focus on longer-term policy
questions until we get beyond this immediate
period of chaos. (FOMC Transcripts, conference
call, October 20, 1987, p. 1)

On September 29, 1998, the FOMC reduced
the funds rate target by 25 basis points. This was
followed by a two additional reductions of 25
basis points on October 15 and November 17.
Chairman Greenspan noted the following at the
February 1999 FOMC meeting:
I have not heard it argued specifically, but our
75 basis point action last fall was directed at
countering a freezing-up of financial markets,
which constituted a demonstrable threat to the
stability of our economy, and arguably we have
largely succeeded. It is true that one can still
observe some residual impact of the liquidity
problems that we have experienced, with yields
on junk bonds remaining significantly above
Treasuries and even obligations rated A and AA
still running spreads against Treasuries that
we haven’t seen for a very long time. If it is
correct that we have succeeded, then one could
argue that we ought to reverse at least part of
our easing moves. (FOMC Transcripts,
February 2-3, 1999, p. 104)

The funds rate target established in November
was maintained until the FOMC meeting in June
1999, though no argument was made that financial
markets remained unsettled after November.
On December 21, 1999, the FOMC press
release noted that the funds rate target was kept
unchanged, in spite of
the possibility that over time increases in
demand will continue to exceed the growth
in potential supply, even after taking account

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

2007

473

Rasche and Williams

of the remarkable rise in productivity growth.
(FOMC Press Release, December 21, 1999)

The maintenance of the existing target funds rate
was explained by concerns about the century date
change:
Nonetheless, in light of market uncertainties
associated with the century date change, the
Committee decided to adopt a symmetric directive in order to indicate that the focus of policy
in the intermeeting period must be ensuring a
smooth transition into the Year 2000. (FOMC
Press Release, December 21, 1999)

On September 17, 2001, the FOMC press
release noted that the funds rate target was
reduced 50 basis points in response to the uncertainty about financial market conditions in light
of the terrorist attack on the World Trade Center:

If the objective of a well-identified model is
achieved, then how should it be used to address
the question of the effectiveness of monetary
policy? Impulse response functions and variance
decompositions that investigate the response to a
monetary shock may not be the most informative
analyses. These address only how the economy
responds to the unpredictable component of
monetary policy—the deviations from rule-like
behavior. Cagan (1989, p. 135) complained that
in the VAR analysis available at the time, the
impact of such residuals was so small as to be
implausible:

It’s clear that the events of last week, at a minimum, have created a heightened degree of fear
and uncertainty that is placing considerable
downward pressure on asset prices, increasing
the probability of an asset price deflation, with
its obvious impact on the economy. (FOMC
Transcripts, conference call, September 17,
2001)

By removing all serial and cross correlations
from economic series, VAR reduces them to
exogenous movements and looks for correlation between these movements in each pair of
series. But these exogenous movements are
little more that isolated blips in the series,
which in monetary growth have little effect
on GNP. The financial system filters out the
effect on monetary blips. Only changes in
monetary growth that are maintained for an
extended period of time affect business activity
These extended changes in monetary growth,
however, exhibit serial correlation and, despite
their variable lags in affecting output and
prices, tend to be correlated with cyclical
movements in other economic variables. The
VAR accordingly eliminates the correlated
movements in money as endogenous to the
economic system. Thus does this technique
give new meaning to the old cliché of “throwing the baby out with the bathwater.”

It is worth noting that, while in real time
FOMC participants were concerned about significant weakness in economic activity in the fourth
quarter of 2001, the current estimate is that gross
domestic product (GDP) grew at a positive 1.6
percent annual rate in that quarter.
Our conclusion from these questions is that
considerable care and additional research is
required to ensure that a valid identified model
of the economy is constructed to enable us to
draw inferences about the effectiveness of monetary policy as a tool for short-run stabilization of
an economy. The number of issues that remain
to be addressed suggests that we are a long way
from a definitive answer.

An alternative investigation is to vary the
parameters in the equation of the identified economic model that characterize the rule-like behavior of the monetary authorities. The question then
becomes not how effective monetary policy has
been in stabilizing the economy under the historical characterization of policy, but how effective
it could be with alternative rule-like behaviors.
Christiano, Eichenbaum, and Evans (1999, pp.
134-46) argue that with VAR models this type of
analysis may be difficult, because identification
of monetary policy shocks is not sufficient to
identify the historical policy rule pursued by the
central bank. The answer to the question of how
effective monetary policy could be in short-run

The Federal Reserve will continue to supply
unusually large volumes of liquidity to the
financial markets, as needed, until more normal
market functioning is restored. (FOMC Press
Release, September 17, 2001)

On the same date, Chairman Greenspan noted:

474

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

stabilization likely depends on the nature of the
shocks that are assumed to hit the economy and,
at least for some shocks, the relative tolerance for
short-run inflation volatility versus output
volatility.
Finally, has increased transparency and
accountability of monetary authorities led to
increased economic stability? This question has
been raised in several contexts. First, some analysts have argued that the “Great Moderation”
since approximately 1983 is substantially due
to better monetary policy and improved transparency. Stock and Watson (2003, p. 29) use
three different econometric models of the U.S.
economy and replace their estimate of a post-1984
monetary policy rule with their estimate of a pre1979 monetary policy rule. They conclude from
these experiments that the models “all suggest
that improved monetary control brought inflation
under control, but accounts for only a small fraction—among the models fit to the United States
data, less than 10 percent—of the reduction in
output volatility.”
Other analysts argue that improved transparency and accountability of central banks
anchor long-term inflation expectations more
firmly, thus giving central banks more latitude to
pursue short-run stabilization objectives. Support
for this argument requires two kinds of research:
(i) What evidence would support the hypothesis
that long-term inflation expectations are less
variable and (ii) has the rule-like behavior of any
central bank become more aggressive in reacting
to short-term fluctuations of output? Levin,
Natalucci, and Piger (2004) provide some evidence on both of these issues by comparing
inflation-targeting industrial countries with
industrial countries that do not announce inflation targets; they conclude (p. 75) that inflation
targeting in these countries has “played a role in
anchoring inflation expectations and in reducing
inflation persistence.”
Chairman Greenspan early on argued that a
low and stable inflation environment contributed
to the higher rate of productivity growth in the
United States after 1995:
Given these real-world uncertainties, it is
important for policymakers to be as explicit

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

as possible about not only the central bank’s
long-run inflation objective but also about its
short-run policy objectives. The more ambiguous policymakers are about these objectives,
the more difficult it will be for the public to
differentiate policy actions that may reflect a
change in the central bank’s long-run inflation
objective from actions intended only to offset
the effects of real shocks on economic activity
…Implicit in that argument, if we are to move
toward price stability, is that the process in and
of itself induces an acceleration of productivity.
(FOMC Transcripts, July 2-3, 1996, p. 47)
It is not that low or stable prices are an environment that is conducive to capital investment
to reduce costs, but rather that it is an environment that forces productivity enhancements.
It forces people who want to stay in business
to take those actions—such as cutting down the
size of the cafeteria, reducing overtime, and
taking away managers’ drivers—that they did
not want to take before in the ordinary course
of business in a modest inflationary environment because it was easier then just to raise
prices to maintain margins. If you force the
price level down, you induce real reallocations
of resources because to stay in business firms
have to achieve real as distinct from nominal
efficiencies. (FOMC Transcripts, July 2-3, 1996,
p. 67)

This is an intriguing hypothesis that is difficult to investigate, given the limited understanding and theory of the determinants of productivity
growth. Unfortunately, it is difficult to reconcile
this hypothesis with the apparent uniqueness of
the U.S. experience with the “productivity boom”
in the face of almost worldwide low and stable
inflation over the past decade.

PROBLEMS IN THE
IMPLEMENTATION OF SHORTRUN STABILIZATION POLICY
One important issue for the implementation
of short-run stabilization policy that did not
receive much attention for a considerable period
of time is the inherent uncertainty of the environment in which central bankers make decisions.
S E P T E M B E R / O C TO B E R

2007

475

Rasche and Williams

There are several dimensions to this uncertainty:
(i) lack of accurate information about the contemporary state of the economy, (ii) inability to forecast accurately the future path of the economy,
and (iii) lack of accurate information about how
policy actions affect the economy.
Two problems face central bankers (and policymakers in general) in assessing the need for a
short-run stabilization action: lags in the availability of data and measurement error in preliminary data.
In the United States, major economic statistics
are available at either monthly or quarterly frequency, usually with an initial publication lag of
a month or two. In other countries, comparable
data may be measured at lower frequency and
with longer publication lags. Consequently, most
formal statistical data that are available for policy
deliberations are “stale.” In the FOMC process,
such data are supplemented by anecdotal data
from the various Federal Reserve Districts.23 The
latter data are not collected from scientific surveys,
and the number of respondents surveyed is small.
Hence, there is a danger of inappropriately extrapolating from the small environment to the macroeconomy. Nevertheless, such reports can give
insights into and reduce, though not eliminate,
uncertainty about emerging trends.
The second problem, measurement error, is
well known; but until recently, it did not receive
much attention, probably because it has been
regarded as a mundane problem and research
into it is unlikely to receive much attention. In
appears that, recently, attitudes have been changing. Research using real-time data has become
more fashionable. Some of this research
(Orphanides, 2001 and 2002) alleges that the
principal culprit in the Great Inflation in the
United States was systematic bias in the real-time
assessment of “potential output” and the “output
gap” in FOMC deliberations. Nevertheless, formal
consideration of measurement error in forecasting
models, whether constructed by private sector
entities or by the staff of policy agencies, remains
underdeveloped, even though the econometric
methodology is well understood. The paucity of
23

See, for example, Poole (2002).

476

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

2007

readily accessible vintage data may contribute to
this problem.24
An additional issue is the limited accuracy
in the forecasts or projections that are available
to monetary policymakers. Absent instantaneous
reaction of the economy to policy actions, effective
stabilization actions require an assessment of the
future state of the economy. Gavin and Mandal
(2001) found the accuracy of the forecasts by
FOMC participants as recorded in monetary policy
reports from 1983 through 1994 for real output
growth are comparable to those of private forecasters (e.g., Blue Chip forecasters).25 However,
the root-mean-squared forecast error at 12- and
18-month horizons was roughly 1 percent (at
annualized rates.) At a 6-month horizon the forecast error was 0.75 percent. In a subsequent
analysis, Gavin and Mandal (2003) extended the
sample of forecasts to 1979-2001. For this longer
sample, they found that the root-mean-squared
forecast errors at the 12- and 18-month horizons
were 1.32 and 1.59 percent, respectively. The
same statistic at a 6-month forecast horizon was
only slightly less than 1 percent.26 This forecast
(in)accuracy suggests that variations in real output growth, from recessions to rapid expansions,
cannot be reliably distinguished on a horizon as
short as a year.
The projection accuracy for real output of the
Reserve Bank of New Zealand (RBNZ) appears to
be comparable to that of the participants in the
FOMC.27 Root-mean-squared projection errors of
24

For the United States, a limited amount of vintage data has been
reconstructed by the research staff of the Federal Reserve Bank of
Philadelphia. Complete archives of the FRED® data base have been
preserved since the web version of this service was introduced in
1996 at least at monthly intervals and since 1999 at weekly intervals. A new data service (ArchivaL FRED, i.e., ALFRED®) has been
implemented by the Research Division of the Federal Reserve
Bank of St. Louis. ALFRED allows the user to retrieve a data list
that is indexed with an “as of” vintage date. Over time, vintage
data that was preserved on hard copy of National Economic Trends
and Monetary Trends will be added to this archive.

25

Gavin and Mandal (2001, Table 2): Forecasts are fourth-quarter
over fourth-quarter growth rates.

26

Gavin and Mandal (2003, Table 1): Forecasts are fourth-quarter
over fourth-quarter growth rates.

27

See Reserve Bank of New Zealand’s “The Projection Process and
Accuracy of the RBNZ Projections”; www.rbnz.govt.nz/monpol/
review/0096577.html.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

the RBNZ are reported as 1 percent at a 1-quarter
horizon and 1.5 percent at a 1-year horizon.
The Bank of England publishes estimates of
the “uncertainty associated with its numeric
projections of inflation and GDP growth” with
each of its Inflation Reports.28 At the 1-year projection horizon conditioned on market interest
rate expectations, the reported uncertainty measure is 0.76 percent; at the 2-year horizon it is 1.0
percent; and at the 3-year horizon it is 1.10 percent. These values are on the order of 50 percent
of the root-mean-squared error of the RBNZ and
FOMC projections at comparable horizons, but
still suggest substantial uncertainty relative to
business cycle fluctuations in real GDP. Other
inflation targeting central banks also make public
projections of real output growth, though this
information does not appear to have a long history and we have not found any other analyses
of the performance of these projections.29
The final problem is the paucity of accurate
information about the dynamic effects of policy
actions. Specifically, received macroeconomic
theories generally provide little insight into
dynamic structures. This is reflected in the VAR
paradigm that eschews any restrictions on
dynamics.
One perspective, associated with Milton
Friedman, is that lags in the impact of monetary
policy are “long and variable.” Another perspective is derived from impulse response functions
of econometric models, including VAR specifications. In many such models, the effect of a shock
to the monetary policy variable is constrained to
be zero as part of the identifying restrictions
imposed on the data. In such models, a typical
response pattern is that several quarters elapse
before a significant response of real output builds
28

29

The recent estimates are in the “Numerical Parameters of Inflation
Report Probability Distributions, February 2005”;
www.bankofengland.co.uk/publications/inflationreport/index.htm.
This report indicates that the uncertainty measure is the standard
deviation of the forecast error in those cases where the distribution
of forecast errors is symmetric.
We have found quantitative projections/forecasts of real output in
published inflation/monetary reports of the central banks of Chile,
Hungary, Iceland, Israel, Korea, Mexico, Norway, Peru, Sweden,
and the United Kingdom. Undoubtedly, we have missed some
reports and we have not completed a tabulation of all published
estimates.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

up, and then this response dissipates over 12 to
18 months.30 In general, estimated confidence
intervals around the impulse response functions
are quite wide. This leaves a policymaker interested in short-run stabilization with a difficult
and unfortunate dilemma: The impact of a policy
action at any horizon is highly uncertain, and
the horizon over which any policy action is most
likely to have a major impact is one where the
future is not predicted with any precision.

CONCLUSION
Several conclusions seem warranted. First,
inflation-targeting central banks appear to have
an admirable record of consistently hitting targets
on a medium-run horizon. However, it is not
clear what the marginal contribution of inflation
targeting beyond a credible commitment to price
stability is, because the Federal Reserve, which
eschews an inflation-targeting framework, has
accumulated a comparable record of low and
stable inflation.
Second, it is not clear what will happen to
low and stable inflation if “bad shocks” are realized and the going gets tough. “Good luck” in the
form of a decade or two of relatively mild shocks
cannot be ruled out as a significant environmental factor during the inflation-targeting period
(see Stock and Watson, 2003, pp. 46-47.)
Finally, the case for consistently effective
short-run monetary stabilization policies is problematic—there are just too many dimensions to
uncertainty in the environment in which central
banks operate.

REFERENCES
AFX Asia. “Santomero: Fed Should Adopt Explicit
Inflation Target.” October 4, 2004.
Anderson, Leonall C. and Carlson, Keith. “A
Monetarist Model for Economic Stabilization.”
Federal Reserve Bank of St. Louis Review, April
1970, 52(4), pp. 7-25.
30

Impulse response functions that are typical of those derived from
VAR analysis can be found in Christiano, Eichenbaum, and Evans
(1999, Figures 2 and 4).

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

2007

477

Rasche and Williams

Anderson, Leonall C. and Jordan, Jerry L. “Monetary
and Fiscal Actions: A Test of Their Relative
Importance in Economic Stabilization.” Federal
Reserve Bank of St. Louis Review, November 1968,
50(11), pp. 11-24.
Bernanke, Ben S. “Alternative Explanations of the
Money-Income Correlation.” Carnegie-Rochester
Conference Series on Public Policy, Autumn 1986,
25, pp. 49-99.
Brunner, Karl and Meltzer, Allan H. “What Did We
Learn from the Monetary Experience of the United
States in the Great Depression?” Canadian Journal
of Economics, May 1968, 1(2), pp. 324-48; reprinted
in A.H. Meltzer, Money, Credit and Policy.
Northampton: Edward Elgar, 1995, pp. 104-18.
Cagan, Phillip. “Money-Income Causality—A Critical
Review of the Literature Since A Monetary History,”
in M.D. Bordo, ed., Money, History and International
Finance: Essays in Honor of Anna J. Schwartz.
Chicago: University of Chicago Press, 1989, pp.
117-51.
Calvo, Guillermo A. “Staggered Prices in a UtilityMaximizing Framework.” Journal of Monetary
Economics, September 1983, 12(3), pp. 383-98.

Journal of Political Economy, February 1977, 85(1),
pp. 191-205.
Federal Open Market Committee. Transcripts,
Minutes, and Press Releases. Various dates.
Federal Reserve Bank of Boston. Housing and
Monetary Policy. Conference Series, Number 4,
October 1970.
Friedman, Milton and Meiselman, David. “The
Relative Stability of Monetary Velocity and the
Investment Multiplier in the United States, 18971958,” in Milton Friedman et al., eds., Stabilization
Policies. Englewood Cliffs, NJ: Prentice-Hall, 1963.
Friedman, Milton and Schwartz, Anna J. A Monetary
History of the United States: 1867-1960. Princeton:
Princeton University Press, 1963.
Fry, Maxwell; Julius, Deanne; Mahadeva, Lavan;
Roger, Sandra and Sterne, Gabriel. “Key Issues in
the Choice of Monetary Policy Framework,” in
L. Mahadeva and G. Sterne, eds., Monetary Policy
Frameworks in a Global Context. London:
Routledge, 2000.
Gavin, William T. and Mandal, Rachel J. “Forecasting
Inflation and Growth: Do Private Forecasts Match
Those of Policymakers?” Federal Reserve Bank of
St. Louis Review, May/June 2001, 83(3), pp. 11-19

Christiano, Lawrence J.; Eichenbaum, Martin and
Evans, Charles L. “Monetary Policy Shocks: What
Have We Learned and to What End?” in J.B. Taylor
and M. Woodford, eds., Handbook of Macroeconomics. Volume 1A. Amsterdam: Elsevier,
1999, pp. 65-148.

Gavin William T. and Mandal, Rachel J. “Evaluating
FOMC Forecasts.” International Journal of
Forecasting, 2003, 19(4), pp. 655-67.

Clarida, Richard; Galí, Jordi and Gertler, Mark. “The
Science of Monetary Policy: A New Keynesian
Perspective.” Journal of Economic Literature,
December 1999, 37(4), pp. 1661-707.

Goodfriend, Marvin “Inflation Targeting in the
United States?” in Bernanke, B.S. and M. Woodford,
The Inflation-Targeting Debate. Chicago:
University of Chicago, 2005, pp. 311-37.

Council of Economic Advisers. Economic Report of
the President. Washington, DC: 1962, 1963, 1975.

Grebler, Leo and Maisel, Sherman J. “Determinants
of Residential Construction: A Review of Present
Knowledge,” in Commission on Money and Credit,
Impacts of Monetary Policy. Englewood Cliffs, NJ:
Prentice-Hall, 1963, pp. 475-620.

Derby, Michael. Dow Jones Commodity Wire report
of Ben S. Bernanke’s remarks at Stanford University,
February 11, 2005.
Fischer, Stanley. “Long-Term Contracts, Rational
Expectations, and the Optimal Money Supply Rule.”

478

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

2007

King, Mervyn. “No Money, No Inflation—The Role of
Money in the Economy.” Bank of England Quarterly
Bulletin, Summer 2002, 42(2), pp. 162-77.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Kohn, Donald L. “Comment,” in B.S. Bernanke and
M. Woodford, eds., The Inflation-Targeting Debate.
Chicago: University of Chicago, 2005, pp. 337-50.
Lacker, Jeffrey M. “Inflation Targeting and the
Conduct of Monetary Policy.” Presented at the
University of Richmond Robins School of Business,
March 1, 2005; www.rich.frb.org/media/speeches/
index.cfm/id=70.
Levin, Andrew T.; Natalucci, Fabio N. and Piger,
Jeremy M. “The Macroeconomic Effects of Inflation
Targeting.” Federal Reserve Bank of St. Louis
Review, July/August 2004, 86(4), pp. 51-80.
Lindsey, David E.; Orphanides, Athanasios and
Rasche, Robert H. “The Reform of October 1979:
How It Happened and Why.” Federal Reserve Bank
of St. Louis Review, March/April 2005, 87(2, Part 2),
pp. 187-236.
Liederman, Leonardo and Svensson, Lars E.O.
Inflation Targets. London: Centre for Economic
Policy Research, 1995.
Loayza, Norman V. and Soto, Raimundo, eds. Inflation
Targeting: Design, Performance, Challenges. Series
on Central Banking, Analysis, and Economic
Policies. Santiago: Central Bank of Chile, 2002.
McCallum, Bennett T. “Price Level Determinacy with
an Interest Rate Policy Rule and Rational
Expectations.” Journal of Monetary Economics,
November 1981, 8(3), pp. 319-29.
McCallum, Bennett T. and Nelson, Edward. “Targeting
vs. Instrument Rules For Monetary Policy.” Working
Paper 2004-011A, Federal Reserve Bank of St. Louis,
June 2004.

Meyer, Laurence H. “Practical Problems and
Obstacles to Inflation Targeting.” Federal Reserve
Bank of St. Louis Review, July/August 2004, 86(4),
pp. 151-60.
Mishkin, Frederic and Schmidt-Hebbel, Klaus. “A
Decade of Inflation Targeting in the World,” in
Inflation Targeting: Design, Performance, Challenges.
Series on Central Banking, Analysis, and Economic
Policies. Santiago: Central Bank of Chile, 2002,
pp. 171-219.
Morande, Felipe. “A Decade of Inflation Targeting in
Chile,” in Inflation Targeting: Design, Performance,
Challenges. Series on Central Banking, Analysis,
and Economic Policies. Santiago: Central Bank of
Chile, 2002, pp. 583-626.
Orphanides, Athanasios. “Monetary Policy Rules
Based on Real-Time Data.” American Economic
Review, September 2001, 91(4), pp. 964-85.
Orphanides, Athanasios. “Monetary Policy Rules and
the Great Inflation.” American Economic Review,
May 2002, 92(2), pp. 115-20.
Orphanides, Athanasios and van Norden, Simon.
“The Unreliability of Output Gap Estimates in Real
Time.” Review of Economics and Statistics,
November 2002, 84(4), pp. 569-83.
Poole, William. “The Role of Anecdotal Information
in Fed Policymaking.” Presented at Washington
University in St. Louis, February 13, 2002;
http://stlouisfed.org/news/speeches/2002/
02_13_02.html.
Poole, William. “FOMC Transparency.” Federal
Reserve Bank of St. Louis Review, January/February
2005a, 87(1), pp. 1-9.

Meltzer, Allan H. ”Monetary and other Explanations
of the Start of the Great Depression.” Journal of
Monetary Economics, November 1976, 2(4),
pp. 455-71; reprinted in A.H. Meltzer, Money,
Credit and Policy. Northampton: Edward Elgar,
1995, pp. 119-35.

Poole, William. “How Should the Fed Communicate?”
Presented at Princeton University, April 2, 2005b;
http://stlouisfed.org/news/speeches/2005/
4_02_05.htm.

Meltzer, Allan H. A History of the Federal Reserve
System, Volume 1: 1913-1951. Chicago: University
of Chicago, 2003.

Radcliffe, G.B.E. “Committee on the Working of the
Monetary System: Report.” London: Her Majesty’s
Stationary Office, August 1959.

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

479

Rasche and Williams

Reuters News. “Fed’s Yellen Says Inflation Target
Has Some Merit.” February 11, 2005.
Romer, Christina D. and Romer, David H. “Does
Monetary Policy Matter? A New Test in the Spirit
of Friedman and Schwartz.” NBER Macroeconomics
Annual 1989, 1989, pp. 121-70.
Romer, Christina D. and Romer, David H. “A
Rehabilitation of Monetary Policy in the 1950s.”
American Economic Review, May 2002a, 92(2),
pp. 121-27.
Romer, Christina D. and Romer, David H. “The
Evolution of Economic Understanding and Postwar
Stabilization Policy.” Proceedings of the symposium, “Rethinking Stabilization Policy” sponsored
by the Federal Reserve Bank of Kansas City, Jackson
Hole, Wyoming, August 29-31, 2002b;
www.kansascityfed.org/Publicat/Sympos/2002/
sym02prg.htm.
Schreft, Stacey L. “Credit Controls: 1980.” Federal
Reserve Bank of Richmond Economic Review,
November/December 1990, 76(6), pp. 25-55.
Sargent, Thomas J. and Wallace, Neil. “‘Rational’
Expectations, the Optimal Monetary Instrument,
and the Optimal Money Supply Rule.” Journal of
Political Economy, April 1975, 83(2), pp. 241-54.
Sims, Christopher A. “Macroeconomics and Reality.”
Econometrica, January 1980, 48(1), pp. 1-48.
Sims, Christopher A. “Are Forecasting Models
Usable for Policy Analysis?” Federal Reserve Bank
of Minneapolis Quarterly Review, Winter 1986, 10,
pp. 2-16.

Changing Environment,” sponsored by the Federal
Reserve Bank of Kansas City, Jackson Hole, Wyoming,
August 28-30, 2003; www.kansascityfed.org/
Publicat/Sympos/2003/pdf/Stockwatson2003.pdf.
Svensson, Lars E.O. “Commentary,” Federal Reserve
Bank of St. Louis Review, July/August 2004, 86(4),
pp. 161-4.
Svensson, Lars E.O. “Targeting versus Instrument
Rules for Monetary Policy: What Is Wrong with
McCallum and Nelson?” Federal Reserve Bank of
St. Louis Review, September/October 2005, 87(5),
pp. 613-25.
Taylor, John B. “Aggregate Dynamics and Staggered
Contracts.” Journal of Political Economy, February
1980, 88, pp. 1-23.
Taylor, John B. “Discretion Versus Rules in Practice.”
Carnegie-Rochester Conference Series on Public
Policy, December 1993, 39, pp. 195-214.
Thornton, Daniel T. and Wheelock, David C.
“A History of the Asymmetric Policy Directive.”
Federal Reserve Bank of St. Louis Review,
September/October 2000, 82(5), pp. 1-16.
Von Hagen, Juergen. “Hat die Geldmenge ausgedient?”
Perspektiven der Wirtschaftspolitik, November 2004,
5(4), pp. 423-53.
Wold, H.O.A. “Causality and Econometrics,”
Econometrica, April 1954, 22(2), pp. 162-77.
Wold, H.O.A. “A Generalization of Causal Chain
Models.” Econometrica, 1960, 28(2), pp. 443-62.

Stock, James H. and Watson, Mark W. “Has the
Business Cycle Changed: Evidence and
Explanations.” Proceedings of the symposium,
“Monetary Policy and Uncertainty: Adapting to a

480

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

481

Rasche and Williams

APPENDIX
Characteristics of Inflation-Targeting Regimes of Countries
Country

Inflation measure

Current target value (%)

Inflation targets (%)

Australia

Headline CPI

2-3

2-3
Treasury underlying CPI
until 1998 and CPI thereafter

Brazil

Headline CPI

4.5 ± 2.5

1999: 8 ± 2
2000: 6 ± 2
2001: 4 ± 2
2002: 3.5 ± 2.
2003: 4 ± 2.52
2004: 4.5 ± 2.5

CPI and core CPI3

2±1

1991: 3-5 (for 22 months)
1992: 2-4 (multiyear)
June 1994: 1.5-3.5
1995-2004: 1-3

Headline CPI

2-4

1991: 15-20
1992: 13-16
1993: 10-12
1994: 9-11
1995: 8
1996: 6.5
1997: 5.5
1998: 4.5
1999: 4.3
2000: ±3.5
2001 onward: 2-4

CPI

6

1999: 15
2000: 10
2001: 8
2002: 6
2003: 4-6
2004: 5-6
2005: 4.5-5.5
2006: 3.5-5.5

Canada

Chile

Colombia

NOTE: 1 Reserve Bank of Australia (RBA). Although the first formal agreement on the conduct of monetary policy between the RBA and
the government was made in 1996, the RBA effectively adopted inflation targeting in 1993. 2 Bank of Brazil Open Letter 2002, 2003,
2004; www.bcb.gov.br/ingles/relinf/OpenLetter2003.pdf. 3 Canada’s core CPI excludes food, energy, and the effect of indirect taxes.

482

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Adoption date

Who sets the target?

Separate inflation report

Published inflation forecast

June 19931

Government and
central bank jointly

No: statements on
monetary policy
contain prospects for
inflation growth

Yes: monetary policy report

June 1999

Government in consultation
with central bank

Yes

Yes

February 1991

Government and
central bank jointly

No

Yes: monetary policy report

September 1990

Central bank in consultation
with government

Yes

Yes

September 1999

Jointly by government
and central bank

Yes

Yes

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

483

Rasche and Williams

APPENDIX, cont’d
Characteristics of Inflation-Targeting Regimes of Countries
Inflation measure

Current target value (%)

Inflation targets (%)

Net inflation
through 20014;
headline CPI thereafter

1-3

Net inflation
1998: 5.5-6.5
1999: 4-5
2000: 3.5-5.5
2001: 2-4
Headline CPI
band starts Jan. 2002: 3-5;
band ends Dec. 2005: 2-4;
from 20065: 3

Finland6

Core CPI

NA

Since 1993: 2

Hungary

CPI

4±1

2001: 7.5 ± 1
2002: 4.5 ± 1
2003: 3.5 ± 1
2004: 3.5 ± 1
2005: 4 ± 1
2006: 3.5 ± 1

Iceland

CPI

2.5 ± 1.5

Since 2001: 2.5 ± 1.5

Headline CPI

1-3

1992: 14-15
1993: 10
1994: 8
1995: 8-11
1996: 8-10
1997: 7-10
1998: 7-10
1999: 4
2000: 3-4
2001: 3-4
2002: 3-4
2003: 1-3
2004: 1-3

Country
Czech Republic

Israel

NOTE: 4 Net inflation is calculated by the Czech Statistical Office as the growth of prices in the unregulated part of the consumer
basket adjusted for changes in indirect taxes and for abolition of subsidies. 5 The 3 percent inflation target had been announced for the
period from January 2006 until the Czech Republic’s accession to the euro area. 6 Since 1998, Finland has been a member of the
European System of Central Banks and the Eurosystem.

484

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Adoption date

Who sets the target?

Separate inflation report

Published inflation forecast

January 1998

Central bank

Yes

Yes

February 1993 to
June 1998

Central bank

No

No

July 2001

Central bank

Yes

Yes

March 2001

Central bank and
government jointly

No: monetary bulletin
contains prospects for
inflation growth

Yes

December 1991

Central bank

Yes

Yes

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

485

Rasche and Williams

APPENDIX, cont’d
Characteristics of Inflation-Targeting Regimes of Countries
Country
Korea

Inflation measure

Current target value (%)

Inflation targets (%)

CPI (1998)
core inflation (after 1999)

2.5-3.5

1998: 9 ± 1
1999: 3 ± 1
2000: 2.5 ± 1
2001: 2.5 ± 1
2002: 3 ± 1
2003: 3 ± 1
Jan. 2004: change to mediumterm (instead of annual)
targeting system;
target for 2004-06: 2.5-3.5

Mexico

Headline CPI

3±1

1999: 13
2000: <10
2001: 6.5
2002: 4.5
2003: 3 ± 1
(in line with trading partners)

New Zealand

Headline CPI7

1-3

1990: 3-5
1991: 2.5-4.5
1992: 1.5-3.5
1993-1996: 0-2
1997-2001: 0-3
2002-present: 1-3

CPI-ATE8

2.5

2001-present: 2.5

Headline CPI

2.5 ± 1

1994: 15-20
1995: 9-11
1996: 9.5-11.5
1997: 8-10
1998: 7.5-9
1999: 5-6
2000: 3.5-4
2001: 2.5-3.5
2002: 2.5 ± 1
2003: 2.5 ± 1
2004: 2.5 ± 1

CPI (they monitor core
CPI measures also)

4.5-5.5

2002: 5-6
2003: 4.5-5.5
2004: 4-5
2005: 4-5

Norway
Peru

Philippines

NOTE: 7 Officially CPI, but the Reserve Bank of New Zealand targets “underlying” or “core” inflation. (CPI: Since December 1997, the
CPI excluding credit services was used; before then, overall CPI was used. In late 1999, mortgage interest rates were removed from the
index.) 8 CPI inflation adjusted for tax changes and excluding energy products.

486

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Adoption date

Who sets the target?

Separate inflation report

Published inflation forecast

April 1998

Government in consultation
with central bank

Yes

Yes

January 1999

Central bank

Yes

Yes

March 1990

Government and
central bank jointly

Yes

Yes

March 2001

Government

Yes

Yes

January 1994

Central bank in consultation
with government

Yes

Yes

January 2002

Jointly by central bank
and government

Yes

Yes

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

487

Rasche and Williams

APPENDIX, cont’d
Characteristics of Inflation-Targeting Regimes of Countries
Inflation measure

Current target value (%)

Inflation targets (%)

Headline CPI

2.5 ± 1

1998: <9.5
1999: 6.6-7.8
2000: 5.4-6.8
2001: <6-8, 3 by 2003
2002: 5 ± 1, <4 by 2003
2003: 3 ± 1
2004: 2.5 ± 1

CPIX9

3-6

2003: 3-6

Spain10

Headline CPI

NA

1996: 3.5-4
1997: 2.5
1998: 2

Sweden

Headline CPI

2±1

Since 1995: 2 ± 1

Switzerland

Headline CPI

<2

Since 2000: <2

Core CPI (excludes raw
food and energy prices)

0-3.5

Since 2000: 0-3.5

Headline CPI11

2

1992: RPIX inflation 1-4
1995: “2.5 or less”
1997: 2.5
Dec. 2003: CPI inflation 2 ± 1

Country
Poland

South Africa

Thailand

United Kingdom

NOTE: 9 The CPI for metropolitan and other urban areas excluding the interest cost of mortgage bonds. 10 Since 1998, Spain is a
member of the European System of Central Banks and the Eurosystem. 11 Since December 2003, CPI inflation became the target,
formerly known as the harmonized index of consumer prices. Prior to that, RPIX inflation was targeted.
SOURCE: Authors’ compilations based on each central banks’ web site discussions on monetary policy, individual monetary policy and
inflation reports of each country, Mishkin and Schmidt-Hebbel (2002), and Morande (2002).

488

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

Rasche and Williams

Adoption date

Who sets the target?

Separate inflation report

Published inflation forecast

October 1998

Central bank

Yes

Yes

February 2000

Government after consultation
with the reserve bank and
the national treasury

Yes

No

November 1994
to June 1998

Central bank

Yes

Yes

January 1993

Central bank

Yes

Yes

January 2000

Central bank

No: monetary policy report,
monetary policy assessment,
and inflation forecast
published quarterly

Yes

May 2000

Government in consultation
with central bank

Yes

Yes

October 1992

Government

Yes

Yes

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W

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

2007

489

490

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

2007

F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W