View original document

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

FEDERAL RESERVE BANK OF ST. LOUIS

Contributing Authors
Ben S. Bernanke

Board of Governors of the Federal Reserve
System
20th Street and Constitution, NW
Washington, DC 20551
ben.s.bernanke@frb.gov

Jon Faust

Division of International Finance
Board of Governors of the Federal Reserve
System
20th Street and Constitution, NW
Washington, DC 20551
jon.faust@frb.gov

Benjamin M. Friedman

Department of Economics
Harvard University
Littauer Center 127
Cambridge, MA 02138
bfriedman@harvard.edu

Dale W. Henderson

Division of International Finance
Board of Governors of the Federal Reserve
System
20th Street and Constitution, NW
Washington, DC 20551
dale.henderson@frb.gov

Otmar Issing

Executive Board, European Central Bank
Kaiserstr. 29
60311 Frankfurt, Germany
jens.kuhn@ecb.int

Donald L. Kohn

Board of Governors of the Federal Reserve
System
20th Street and Constitution, NW
Washington, DC 20551
donald.kohn@frb.gov

Kenneth N. Kuttner

Department of Economics
Oberlin College
Rice Hall
10 North Professor Street
Oberlin, OH 44074
kenneth.kuttner@oberlin.edu
Andrew T. Levin

Division of Monetary Affairs
Board of Governors of the Federal Reserve
System
20th Street and Constitution, NW
Washington, DC 20551
andrew.levin@frb.gov
Laurence H. Meyer

Center for Strategic and International Studies
1800 K Street, NW, Suite 400
Washington, DC 20006
lmeyer@egsis.org
Fabio M. Natalucci

Division of International Finance
Board of Governors of the Federal Reserve
System
20th Street and Constitution, NW
Washington, DC 20551
fabio.m.natalucci@frb.gov
Monika Piazzesi

Graduate School of Business
University of Chicago
1101 East 58th St.
Chicago, IL 60637
monika.piazzesi@gsb.uchicago.edu
Jeremy M. Piger

Federal Reserve Bank of St. Louis
P.O. Box 442
St. Louis, MO 63166
jeremy.m.piger@stls.frb.org
Stephanie Schmitt-Grohé

Department of Economics
Duke University
P.O. Box 90097
Durham, NC 27708
grohe@duke.edu

J U LY / AU G U S T 2 0 0 4

1

REVIEW
Lars E.O. Svensson

Department of Economics
Princeton University
111 Fisher Hall
Princeton, NJ 08544
svensson@princeton.edu
Harald Uhlig

Institute for Economic Policy
School of Business Administration and
Economics
Humboldt University
Spandauer Str. 1
10178 Berlin, Germany
uhlig@wiwi.hu_berlin.de
Michael Woodford

Department of Economics
Princeton University
111 Fisher Hall
Princeton, NJ 08544
woodford@princeton.edu

2

J U LY / AU G U S T 2 0 0 4

Editors’ Introduction
Jeremy M. Piger and Daniel L. Thornton

T

hese conference proceedings consist of
extremely thoughtful and provocative papers
and discussions on critical and ancillary
issues that constitute the essence of the inflation
targeting (IT) debate—including a provocative discussion of IT by three of the world’s top policymakers. The breadth and depth of the analysis
by these conference participants is remarkable—
virtually no relevant stone of the IT debate is left
unturned. We cannot possibly do justice to the
participants’ contribution in this introduction. With
this proviso, we first discuss several of the most
important lessons we learned from the conference
papers and discussions and their possible implications for future research. We then discuss what our
prominent panel of policymakers has to say about
IT, especially with respect to these issues. We
emphasize that our brief introduction cannot adequately reflect all of the subtleties and nuances of
the positions taken; our summary is a poor substitute for reading the entire conference proceedings.

Nutters, NETers, LETers, Flexible ITers,
Super Flexible ITers, etc.
Perhaps the most recurrent theme in the conference proceedings is the extent to which IT provides
room for, or alternatively limits, policymakers’ discretion to pursue other objectives—most frequently,
stabilizing the real economy. Faust and Henderson’s
thoughtful and systematic analysis of whether IT
constitutes “best practice” monetary policy addresses this issue directly. They note that the IT perspective is bolstered by the “now nearly universally
accepted” fact that there is “no long-run Phillips
curve trade-off of the traditional variety” (p. 120).
They then distinguish between those who believe
there is no exploitable trade-off between inflation
and the output gap in the short run—NETers—and
those who believe that there is a “limited” exploitable

trade-off—LETers. While these views are distinct,
Faust and Henderson note that “it is sometimes difficult to tell which view various parties take” (p. 122).
In part, this is because some LETers believe that
the degree of exploitability is “quite low.” It is also
because “virtually everyone agrees that demand
shocks push us toward a singular economy perspective” (p. 122). That is, in many standard models, the
appropriate policy response to demand shocks for
NETers and LETers is isomorphic, or nearly so. The
agreement with respect to demand shocks substantially narrows the important difference between
NETers and LETers. Faust and Henderson suggest
that one’s view about how policy should respond
to supply shocks provides a “litmus test for deciding
whether one is in the NET or LET camp” (p. 122).
While they do not address the question directly,
their analysis seems to suggest that all NETers are
likely to be ITers. On the other hand, not all ITers
are NETers. Indeed, they suggest that “many, if not
most, advocates of the ITF [inflation-targeting framework] are LETers” (p. 121).
Faust and Henderson note that, in their lexicon,
NETers are not what Mervyn King has termed
“inflation nutters”—those who consider inflation
the sole policy objective—because NETers believe
that, given the economic structure, stabilizing prices
is “the best we can hope for” (p. 121). In commenting
on Faust and Henderson, however, Ben Friedman
notes that “for practical purposes…these two positions [NETers and nutters] are isomorphic” (p. 147).
From this discussion, we conclude that nutters are
LETers who believe that the central bank should
pursue other objectives only in rare circumstances—
nutters fail the Faust-Henderson’s NETer litmus test.
It is clear from these proceedings that the extent
to which there is an exploitable short-run trade-off
between inflation and the output gap—the extent
to which the NETer view of policy is correct—is an
important issue of the IT debate. Faust and Henderson

Jeremy M. Piger is an economist and Daniel L. Thornton is a vice president and economic advisor at the Federal Reserve Bank of St. Louis.
Federal Reserve Bank of St. Louis Review, July/August 2004, 86(4), pp. 3-13.
© 2004, The Federal Reserve Bank of St. Louis.

J U LY / A U G U S T 2 0 0 4

3

Piger and Thornton

note that the NETer view can arise in models where
the economy is enormously complex—Milton
Friedman and Robert Lucas—or “fortuitously”
simple—Rotemberg and Woodford (1997), King and
Wolman (1999), and Goodfriend and King (2001).
While NETers, per se, appear to be rare, theoretical
analyses indicate that the “tension” between inflation and output stabilization is much smaller than
previously thought. In his survey of the theoretical
literature on “optimal monetary policy,” Woodford
concludes
recent literature on the welfare consequences of alternative monetary policies
finds that there is less tension between inflation stabilization and properly defined real
stabilization objectives than traditional (nonwelfare-theoretic) literature on monetary
stabilization policy has often suggested. It
is not a bad first approximation to say that
the goal of monetary policy should be price
stability. (p. 23)
Woodford notes that there are some instances
where deviations from perfect price stability is optimal policy. For example, it is not optimal in instances
where complete price stability is infeasible—where
the natural rate of interest is temporarily negative or
not efficient—e.g., where the existence of frictions
make it desirable to have the nominal rate more
stable than the natural rate. Woodford notes that
while interest rates are smoothed considerably
under optimal policy in these circumstances, “this
does not require too much variation in inflation”
because the variance trade-off is “quite flat near
the extreme of full inflation stabilization” (p. 24).
Woodford also notes that full stabilization of
prices is not generally optimal in models where
prices of different commodities are impacted differentially by economic shocks. He notes, however,
that “as long as the price index to be stabilized is
appropriately chosen, complete stabilization of a
price index is found (in calibrated models) to be
nearly optimal” (p. 25). Not surprisingly, an analogous
conclusion arises in models where wages are as sticky
as or more sticky than prices. Likewise, the existence
of market power or the existence of distorting taxes
creates situations where “it will not be possible to
simultaneously stabilize inflation and the welfarerelevant output gap.” Nevertheless, Woodford notes
that even in such circumstances,
4

J U LY / A U G U S T 2 0 0 4

REVIEW
the degree of variability of inflation under
an optimal policy may be quite modest…
because the relative weight that should be
placed on the goal of output stabilization,
relative to the weight on inflation stabilization, may not be large. (p. 26)
In her discussion of Woodford’s paper, Stephanie
Schmitt-Grohé notes that most of the theoretical
work that Woodford surveys uses dynamic, stochastic general equilibrium models that require a
“simplifying assumption” to obtain “an analytical
characterization of optimal policy” (p. 43). She then
reports on the findings of some recent work using
“recent advances in computational economics…
that make it feasible and simple to compute higherorder approximations to the equilibrium conditions
of a general class of large stochastic dynamic general
equilibrium models” (p. 43). In a model that allows
for sticky prices, money demand, distortionary taxes,
and a role for fiscal variations in the price level, she
finds that, when the model is calibrated to the U.S.
economy, the mean and standard deviation of inflation for optimal policy “is close to zero”—the NETer
view is approximately correct (p. 44). Moreover,
she reports that “the inflation-volatility tax-ratevolatility trade-off is resolved in favor of inflation
stability not only for degrees of price stickiness
observed in the U.S. economy, but also for much
lesser degrees of price stickiness” (p. 44). She then
notes that this conclusion is robust to a model with
capital accumulation.
Despite the growing theoretical evidence in favor
of inflation stabilization, the tension between inflation stabilization and output stabilization remains
a central issue in the IT debate. In his excellent discussion of practical problems and obstacles to inflation targeting, Larry Meyer characterizes the issue
by making a distinction between what he terms
inflation targeting and inflation targets. He notes
that inflation-targeting countries generally operate
with both an explicit numerical inflation target and
a hierarchical mandate, under which “central banks
are restricted in pursuing other objectives unless
price stability has been achieved” (p. 151, emphasis
added). In contrast, under a dual mandate “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, emphasis added).
Meyer favors an explicit inflation target as part
of a dual mandate, but opposes an inflation-targeting

FEDERAL R ESERVE BANK OF ST. LOUIS

regime with a hierarchical mandate. Meyer believes
that central banks can control the long-run average
rate of inflation—“with respect to inflation, the buck
literally does stop at the central bank” (p. 152). But
his desire for a dual mandate stems from the belief
that “at the margin,” central banks “can damp movements in output around its potential level.” It is less
clear, however, whether he believes that stabilizing
output around its potential level requires a sacrifice
of inflation stabilization, or whether he adheres to
what he describes as Chairman Greenspan’s “unique
vision” of monetary policy. According to Meyer,
Greenspan believes that synergies exist “between
the two objectives for monetary policy—price stability and damping fluctuations around full employment” (p. 153). He notes that Chairman Greenspan
“is generally viewed as being a hawk when it comes
to containing inflation and a dove when it comes to
quickly providing support for a weakening economy”
(p. 153). Whatever his belief, Meyer argues that an
inflation-targeting regime with a hierarchical mandate is a non-starter for the United States—there is
“no chance that the Congress would accept a regime
with a hierarchical mandate that raised the profile
of price stability and diminished the responsibility
of the FOMC for stabilization policy” (p. 154).
In his discussion of Meyer’s paper, Lars Svensson
argues that the hierarchical/dual mandate distinction
is not useful. He notes that with quadratic loss functions of the type that are frequently used by economists to represent the central bank’s objective
function, the inflation target is a choice variable, but
the output target is not—“the output target is subject
to estimation, but it is certainly not subject to choice”
(p. 161). Consequently, he suggests that there is only
a hierarchical mandate for long-run inflation. Consistent with Woodford’s interpretation, he suggests
that the dual mandate is reflected in the choice of
the weight that policymakers give to inflation stabilization relative to output stabilization:
Since all inflation targeters are flexible inflation targeters, in the sense that they are concerned about stability of the real economy…
we can, if we like, talk about inflation targeters as having a dual mandate. But, as long as
we know that we are talking about different
verbal descriptions of monetary policy loss
functions of the kind stated above, I do not
find the dual/hierarchical mandate distinction helpful. (p. 162, emphasis added)

Piger and Thornton

Taxonomy aside, the important question in the
IT debates is: How much inflation stability are policymakers willing to give up to gain some quantity X
in the stability of output? Meyer recognizes that
inflation-targeting regimes have become more
flexible over time, but he believes that the hierarchical mandate nevertheless imposes an inappropriate
and unnecessary constraint on the flexibility of
monetary policymakers as they try to balance their
objectives of price stability and full employment.
From this perspective, this critical issue in the IT
debate appears to hinge on the assumption that
policymakers must give up one thing to get another.
We now turn to this aspect of the conference
proceedings.

Transparency, Accountability, Credibility,
etc.
There appears to be widespread agreement
among the conference participants that the essential
elements of IT are (i) an explicit long-run inflation
objective and (ii) a commitment to transparency.
The latter is closely linked to credibility and, most
importantly, the belief that having a credible longrun inflation objective generates a “win-win” situation for the central bank—a commitment to low
and stable inflation not only makes conducting
monetary policy easier but simultaneously results
in better stabilization outcomes.
Woodford notes that, in some models, uncertainty about the long-run inflation rate “worsens the
trade-off between inflation variability and outputgap variability that is available to the central bank”
(p. 18). Friedman makes the identical point (p. 146).
The idea that low and stable inflation has desirable
consequences for the real economy seems to be at
the heart of the “synergies” that Meyer indicates
are a hallmark of Greenspan’s vision of monetary
policy. In commenting on Levin, Natalucci, and Piger,
Uhlig reaches a similar conclusion. Specifically, he
compares 10-year standard deviations of inflation
and real output growth for the United States over
the period 1940-2002 and concludes that “these
figures seem to suggest that an environment of low
and stable inflation helps to reduce output volatility
and support economic activity” (p. 85).
It is certainly the case that there is some level
of inflation variability where the commitment to a
long-run inflation objective is not credible. If this is
so and if there is a significant improvement in the
inflation/output variability trade-off associated with
J U LY / A U G U S T 2 0 0 4

5

Piger and Thornton

a credible inflation target, it is reasonable to expect
ITers to have a hierarchical mandate, at least, as
Meyer defines it. The hierarchical mandate will be
a natural consequence of the fact (or at least belief)
that thresholds exist for the mean inflation rate and
inflation variability, such that a policy which allows
these measures to stay consistently above these
thresholds generates less-favorable results for things
that policymakers and the public care about, e.g.,
the real economy.
There are, of course, other reasons for transparency. Faust and Henderson note that Lucas
pointed out that “what constitutes optimal policy
is inextricably linked with public expectations about
policy” (p. 122). In addition, noting that the overnight interest rate is of “negligible importance for
economic decisionmaking,” Woodford states “not
only do expectations about policy matter, but, at least
under current conditions, very little else matters”
(p. 16). Moreover, he notes that “actual changes in
overnight rates required to achieve the desired
changes in incentives can be much more modest
when expected future rates move as well” (p. 17).
Hence, according to Woodford the effectiveness of
policy depends critically on the central banker’s
ability to “manage expectations.”
Transparency and the Practice of IT. Assuming that the hallmark of transparency is honesty,
there appears to be concern that some transparent
(and hierarchical) central banks are not being
honest. Meyer relates a story where he was taken
aside by “two of the leading central bankers in the
world” and told that “good central bankers never
admit they pursue stabilization policy” because
“such an admission would reduce the confidence
of the public in your commitment to price stability
and therefore undermine your credibility and effectiveness as a monetary policymaker” (p. 152). He
expressed confusion that the “way to build credibility was to lie…” (p. 152).
Faust and Henderson address this point noting
that a “folk wisdom” of central banking is that
“central banks should ‘do what they do, but only
talk about inflation’ ” (p. 132). Expressing concern
analogous to Meyer’s, they suggest that perhaps “we
should stop praising the ITF and other central banks
for their commitment to transparency; instead we
should lament the fact that central banks cannot
publicly discuss the pursuit of their multiple mandates” (p. 132).
Friedman underscores this point, but adds a
concern. Arguing that “language matters,” Friedman
suggests that
6

J U LY / A U G U S T 2 0 0 4

REVIEW
it is not too great a leap to conjecture that
one consequence of constraining the discussion of monetary policy to be carried out
entirely in terms of an optimal inflation
trajectory will be that concern for real outcomes will atrophy, or even disappear from
policymakers’ consideration altogether. Nor
is it unreasonable to suppose that the hope
that this eventuality will ensue is, for some
advocates, a motivation for favoring inflation
targeting in the first place. (p. 148)
While it is possible that there is an ideological
battle between the near-NETer-LETers and other
LETers, we are inclined to believe that the real issue
is a lack of agreement in the profession—and particularly among policymakers—about the circumstances
where (i) monetary policy can effectively stabilize
output and (ii) the costs of such stabilization actions.
To illustrate this point, we note that in his discussion
of Meyer’s hierarchical/dual mandate distinction,
Svensson argues that it is particularly “misleading
to say that inflation targeters have a hierarchical
mandate but the Fed does not” (p. 162). Svensson
argues that by inserting “sustainable” before
“employment” in the mandate as expressed in the
Federal Reserve Act—to promote effectively the goals
of maximum employment, stable prices, and moderate long-term interest rates—the Fed’s mandate can
be interpreted as describing a loss function where
some positive, but unstated weight is given to stabilizing the output gap. He argues that this is not “substantially different” from the Reserve Bank of New
Zealand’s Policy Target Agreement, which suggests
that in pursuit of its price stability objective, it shall
“seek to avoid unnecessary instability in output,
interest rates and the exchange rate” (p. 162). We
note, however, that Meyer suggests that he believes
that Chairman Greenspan “also believes that low,
stable inflation contributes to strong productivity
growth and hence to a higher maximum sustainable
rate of economic growth,” providing yet “another
reason why maintaining low stable inflation has
significant payoffs for economic performance”
(p. 153). From this perspective, the Fed’s mandate
need not be inconsistent with that of a central banker
who gives no weight to stabilizing output in the loss
function. Meyer is quick to point out that he believes
“that the other members of the FOMC have less
faith in this principle than the Chairman” (p. 153).
Our point is not that the FOMC’s stated mandate is
hierarchical or dual or that Chairman Greenspan is

FEDERAL R ESERVE BANK OF ST. LOUIS

a NETer and not a LETer (we simply don’t know).
Rather, our point is that this example and much of
the analysis and discussion by conference participants suggests that a core issue (if not the core issue)
in the IT debate is the extent to which and the precise circumstances under which central banks can
and should stabilize output. Because this requires
responses to “supply shocks” rather than demand
shocks, it would seem that specific analyses of the
types of supply shocks that central bankers should
respond to and the potential costs associated with
such responses would be a good place to start.
In any event, there is agreement among
Bernanke, Faust and Henderson, Meyer, Svensson,
Uhlig, and Woodford that, in Woodford’s words, “it
would be desirable for central banks to commit themselves to the pursuit of explicit target criteria that
involves real variables as well as inflation” (p. 27).
Woodford notes that this would “increase transparency, facilitating the public’s ability to correctly
anticipate future policy” and, echoing a point made
by Meyer, “help to dispel some of the resistance to
the adoption of inflation targeting in countries like
the United States” (p. 27).

What’s the Advantage of IT?
Faust and Henderson argue that “the first goal
of best-practice [monetary policy] is to get mean
inflation right” (p. 124, emphasis added). They then
note that a purpose of IT is to “anchor long-run
inflation expectations,” by leaving “little room for
misunderstanding this objective.” That being said,
it is natural to ask: Why must the long-run inflation
objective be explicit? Why can’t it be fuzzy? The
unquestionable existence of “control error” suggests
that inflation targeting is inherently fuzzy from the
public’s perspective. This is illustrated in Faust and
Henderson’s discussion of best-practice monetary
policy. They note that many IT central banks have
“target ranges” for the IT. Under their interpretation
of best-practice monetary policy, “a target range is
purely descriptive in that it states that inflation will
be within the range π*± θ [the inflation target plus/
minus a positive fixed amount] most of the time”
(p. 125). They note that “no incident of inflation
crossing the boundary is evidence of central bank
misbehavior; only excessive frequency of being outside the interval constitutes such evidence” (p. 125).
They note too that “excessive frequency of being
inside the range is also evidence of misbehavior”
for a central bank that professes to have other noninflation policy objectives (p. 125). The point is that

Piger and Thornton

given such an interpretation, it may take the public
some time (with the length of time determined in
part by the nature of the control errors, assuming
their existence) to determine whether the central
bank is in fact attempting to achieve its stated inflation objective or is misbehaving with respect to it.
Meyer suggests that having an explicit inflation
objective should (i) “contribute to anchoring inflation
expectations,” (ii) “improve the coherence of the
deliberations and the policy outcomes” by ensuring
a consensus among policymakers about the desired
long-term inflation rate, and (iii) “enhance the ability
of bond market participants to anticipate the future
course of monetary policy…thereby improving the
effectiveness of policy” along the lines discussed by
Woodford (p. 154).
Levin, Natalucci, and Piger (LNP) attempt to
assess on empirical grounds whether or not inflation
targeting has made a difference for economic performance. In the tradition of the existing literature
on this topic, LNP ask whether the economic performance of a group of countries that have adopted
IT differs substantially from a group of countries that
have not. LNP first investigate whether the data are
consistent with the claim that inflation expectations
are anchored by inflation targeting. Using privatesector inflation forecasts for industrial countries
obtained from surveys, LNP find that the unconditional volatility of these inflation forecasts is not
noticeably different between IT and non-IT economies. Digging deeper however, LNP find that the
source of volatility does seem to differ. In particular,
LNP find that for non-IT economies, long-run inflation forecasts exhibit a highly significant correlation
with past realized inflation. In contrast, this correlation is largely absent for the IT countries. LNP argue
that this is suggestive evidence that the inflationtargeting central banks have been quite successful
in delinking expectations from realized inflation. It
is also consistent with the claim that IT anchors
inflation expectations.
LNP also investigate whether there is any difference in the dynamics of actual inflation between
IT and non-IT economies. Again, they find no evidence that inflation has been unconditionally less
volatile in IT vs. non-IT economies. However, they
again find evidence that the source of this volatility
differs. In particular, shocks to core CPI inflation are
substantially less persistent in the group of IT economies than in the non-IT economies. In particular,
it is difficult to reject the hypothesis that core CPI
inflation follows a random walk in most of the nonJ U LY / A U G U S T 2 0 0 4

7

Piger and Thornton

IT economies, whereas this hypothesis can be easily
rejected for most of the IT economies. LNP note that
this suggests that inflation dynamics in the non-IT
economies display a substantial propagation component that augments shocks, whereas inflation
dynamics in the IT economies quickly revert to the
mean following a shock.
LNP also consider a group of emerging-market
economies who have adopted IT more recently. For
these recent adopters they confirm the result, found
by others for industrial IT economies, that the adoption of inflation targeting does not lead to an immediate reduction in long-term inflation expectations.
In this sense, the true benefits of IT may become
apparent only after the regime has gained sufficient
credibility.
In discussing their work, Harald Uhlig does not
question their findings, noting that “The correlations
found by the authors seem to be there, and they can
be read as a list of interesting differences between
countries that have formally adopted inflation targeting and those that have not” (p. 82). He does, however, question their interpretation of these results.
He questions LNP’s division of countries into ITers
and non-ITers, noting that the Bundesbank is classified as a non-ITer by LNP. He notes that (i) it is widely
accepted that the Bundesbank ignored violations of
its money growth targets “if it helped in pursuing
some other, important goal—most notably price
stability,” (ii) the money growth targets were derived
in large part to be consistent with the Bundesbank’s
“underlying goal regarding the desired rate of inflation,” and (iii) the public debates about monetary
policy “were practically always in terms of inflation”
(p. 82).
Uhlig then asks, even if IT has anchored inflation
expectations, “how helpful has that been for the
variables that we ultimately care about?” Noting
that the standard deviation of inflation has been
higher for ITers than non-ITers and that output
growth variability has been essentially the same,
Uhlig suggests that “based on these numbers alone,
one certainly would not want to make the case that
adopting inflation targeting is a good idea” (p. 83).
Acknowledging that his unconditional perspective
may be misleading because ITers may have been
hit by larger shocks or started with less favorable
initial conditions, Uhlig concludes “much more work
than is in this paper is required before it is possible
to conclude that inflation targeters have been more
successful in containing shocks hitting the economy
than nontargeters have been” (p. 83).
8

J U LY / A U G U S T 2 0 0 4

REVIEW
Uhlig also suggests a reinterpretation of LNP’s
finding that inflation volatility has been more persistent for non-ITers. Noting that in models of the
New Keynesian variety,
low-frequency volatility of inflation is OK,
but high-frequency volatility is bad for the
economy and leads to an overall lower level
of economic activity. If this is what would
happen with inflation targeting, which seems
to be what the empirical results suggest, then
this would be an argument against inflation
targeting, not for it. (p. 84)
Finally, Uhlig notes that IT has frequently arisen
endogenously—it occurs “when the economic situation was sufficiently bad” (p. 84). He suggests that
this not only accounts for the failure to adopt inflation targeting by the United States and the European
Monetary Union, but requires that analyses of the
effectiveness of IT must account for fiscal considerations and other structural reforms—“The interesting
question remains whether inflation targeting has
contributed above and beyond fiscal consolidation
or general institutional reforms” (p. 84).

IT and the Rules Versus Discretion
Debate
Faust and Henderson effectively ask: Is IT best
“viewed as a ‘rule’ or the exercise of ‘discretion’ ”?
They note that ITers appear to have different views
about that, but that these differences appear to be
definitional, so that, in their view, IT is “a typical
example of discretion in the classical sense” (p. 119).
Nevertheless, it is clear that many if not most ITers
consider IT to be a rule in the modern sense of a
policy rule, e.g., a Taylor rule. That is, IT is a rule as
defined by Woodford,
a rule under which the central bank’s commitment is defined by a target for certain
variables at a certain distance in the future,
together with a commitment to organize
deliberations about policy actions around
the question of whether the contemplated
actions are consistent with the target. (p. 20)
Moreover, many ITers, such as Woodford, argue
that a commitment to such a policy rule has important consequences for the effectiveness of monetary
policy.

FEDERAL R ESERVE BANK OF ST. LOUIS

In his careful analysis of “The Role of Policy
Rules in Inflation Targeting,” Ken Kuttner notes that
the view of IT as “some sort of a monetary policy
rule” stems, in large part, from the fact that the
adoption of IT by many central banks and the “explosion of research on monetary policy rules” occurred
at much the same time. Consistent with Woodford’s
definition, Kuttner notes that conditional rules “allow
the policymaker to respond in a reasonable (or even
optimal) manner to economic conditions” (p. 90).
Before assessing “empirically the extent to which
IT can be described in terms of simple monetary
policy rules” (p. 89). Kuttner undertakes a detailed
and thoughtful analysis of important issues in the
policy rule debate—ad hoc versus optimal policy
rules, instrument rules versus targeting rules, rules
describing outcomes versus rules based on commitment, and mechanical rules versus guidelines.
After a careful analysis of the “connection
between various definitions of a policy rule…and IT
as it is actually practiced” (p. 92), Kuttner undertakes
a careful empirical analysis of “how well the behavior of IT central banks can be characterized by simple
policy rules” (p. 94). Specifically, he estimates policy
rules for three IT central banks, the Reserve Bank
of New Zealand, the Bank of England, and Sweden’s
Riksbank, and for the Federal Reserve. The unique
twist in Kuttner’s analysis is the “use of the central
bank’s own published inflation and output forecasts,
rather than econometric proxies for the relevant
expectations” (p. 95). Indeed, the three central banks
were chosen because they have “the longest track
record of published, quantitative forecasts” (p. 95).
Kuttner notes that these forecasts (i) reduce the data
requirements and alleviate the need for two-stage
GMM estimation, (ii) “are likely to be more reliable
that those based on simple econometric models,
and (iii) should “embody appropriate assumptions
about the central banks’ intended policy actions”
(p. 95).
After noting some technical differences in the
inflation target and forecasting by each of the IT
central banks and the fact that “an explicit reaction
function or instrument rule does not figure prominently in any of the three central bank’s official
publications” (p. 96), Kuttner begins by estimating
a standard “Taylor rule” with a lagged interest rate
term. Because none of the central banks, save the
Reserve Bank of New Zealand, publishes an estimate
of the output gap, Kuttner estimates the output gap
under some reasonable assumptions. Kuttner finds
that the estimates suggest that this simple policy rule
“is a poor description of policy for all four central

Piger and Thornton

banks” (p. 99). The coefficient on inflation is significant only for Sweden, while the output gap coefficient was significant for three countries, but had
the wrong sign for two—Sweden and the United
Kingdom.
Estimates from a forward-looking Taylor rule
fared better. All coefficients were correctly signed
and statistically significant for New Zealand and
the Taylor principle—defined as the long-run
response of the nominal rate to inflation being
greater than 1—was satisfied for all countries but
the United Kingdom.
Because the estimates of the coefficient on
the lagged interest rate were large, implying “an
extremely high degree of interest rate smoothing,”
Kuttner investigates the possibility that this reflected
“the omission of highly serially correlated variables
from the instrument rule,” possibly corresponding
to the unobserved “judgment” terms emphasized in
Svensson (2003), “rather than interest rate smoothing per se” (p. 101). He does this by including a
variable that captures the “ ‘news’ contained in the
revisions in expectations embodied in the central
banks’ inflation and output forecasts” (p. 101). He
finds that the results are more encouraging than for
either of the previous specifications, but like the
previous specifications, the results are best for New
Zealand and Sweden. Importantly, Kuttner notes that
the estimates support Rudebusch’s (2002) “contention that at least some of the serial correlation in
conventionally specified instrument rules represents
a response to an omitted variable” (p. 103). He argues
that the results demonstrate that IT central banks
exercise judgment in setting policy. He notes that
this “interpretation is particularly clean in the case
of New Zealand,” suggesting that “even the (arguably)
most rule oriented of all the IT central banks apparently still exercises a great deal of judgment in setting policy” (p. 104).
Finally, Kuttner investigates the usefulness of
the policy-rule framework for analyzing policy by
estimating the correlation between the forecast
inflation gap and the forecast change in the output
gap. Kuttner shows that under pre-commitment
(history dependence) this correlation will be positive
“under the assumption of a backward-looking inflation process” and negative if the inflation process
is forward looking. The correlation is estimated at
four-quarter and eight-quarter horizons. In all but
one case the correlation is negative, and the positive correlation is not statistically significant. The
results under discretion were even less supportive
for forward-looking policy. Kuttner notes that his
J U LY / A U G U S T 2 0 0 4

9

Piger and Thornton

generally negative results “will come as no surprise
to those familiar with the practice of IT.” He also
notes that “the lack of a sharp, qualitative difference
between the Fed’s behavior and that of the inflation
targeters will probably do little to alter the priors of
skeptics…who contend the policy makes little practical difference” (p. 107).
In her discussion, Monika Piazzesi applauds
Kuttner’s innovative use of central bank forecasts
to estimate policy rules. She gives two reasons for
being excited about looking at central bank projections that Kuttner did not mention. First, she notes
that there is a great deal of uncertainty about the
appropriate inflation measure to use. She observes
that using central bank projections eliminates this
problem and a host of others—if a central bank
publishes projections of inflation and output “for
different horizons, k, the only question is what horizon k to pick to estimate the policy rule. But picking
k seems easy compared with the host of other problems that we run into otherwise” (p. 114). She then
notes that if central bank projections of these variables are the “right” measure of the current belief
of the central bank about these variables, “we may
be able to estimate policy rules that are stable functions of the projection data” (p. 114).
She also suggests that projections might be used
to investigate “what a model of the economy that
gives rise to these projections would look like.” For
example, she suggests that it would be useful to know
if these projections are biased, if they can be forecast
with macroeconomic variables, if the projection
errors are serially correlated, etc. She then suggests
that the projection data might be used to distinguish
whether the central bank has private information
that is used in making policy decisions, and the
extent to which central bank information is reflected
in financial markets.
Pointing out that the Fed’s Green Book forecasts
are not necessarily the FOMC’s forecasts, she cautions that “projection numbers may not be the numbers that ultimately influence policy decisions”
(p. 115). Finally, she observes that central banks
“may have incentives to distort there projection
numbers,” noting that inflation projections for IT
central banks may play roles similar “to earnings
projections by private firms” (p. 115).

Policymakers’ Views on IT
While our panelists share the view that maintaining low and stable inflation has beneficial
consequences for the real economy, they differed
10

J U LY / A U G U S T 2 0 0 4

REVIEW
considerably on the desirability of IT. Of our three
panelists, the most pro-IT was Ben Bernanke. In an
apparent endorsement of Greenspan’s belief in
synergies between inflation and economic stabilization, Bernanke argued that “the Fed has built strong
credibility as an inflation-fighter…and that this
credibility has allowed the Fed to be relatively flexible
in responding to short-run disturbances to output
and employment without destabilizing inflation
expectations” (p. 165). Despite these gains, Bernanke
suggests “an incremental move toward inflation
targeting” that “might help the Fed communicate
better and perhaps improve policy decisions as well”
(p. 165).
Specifically, Bernanke proposed that the FOMC
announce a long-run inflation rate (OLIR)—the
inflation rate the FOMC believes “achieves the best
average economic performance over time with
respect to both the inflation and output objectives”
(p. 166). Recognizing that one of the impediments
to adopting IT by the United States is the concern
(expressed by several conference participants) that
such a move would reduce the flexibility of policy
to pursue other economic stabilization objectives,
Bernanke recommends that in introducing the OLIR
the FOMC should state that
the FOMC regards this inflation rate as a
long-run objective only and sets no fixed
time frame for reaching it. In particular, in
deciding how quickly to move toward the
long-run inflation objective, the FOMC will
always take into account the implications for
near-term economic and financial stability.
(p. 167)
Arguing that announcing the OLIR would not
have “significant costs,” Bernanke addresses the
question: What are the benefits? Bernanke suggests
five. First, “the announcement of the OLIR should
serve as a useful clarification of the long-run objective of the Fed and would thereby provide a long-run
‘anchor’ to monetary policy.” Second, “the OLIR
should help participants in financial markets price
long-term bonds and other financial assets more
efficiently.” Third, establishing the OLIR should “help
to lower inflation risk in financial markets and other
forms of contracting.” Fourth, consistent with the
synergies vision, it will “tend to stabilize long-term
inflation expectations more broadly, which in turn
would make short-run stabilization policy more
effective.” Finally, Bernanke notes that “although
the announcement of the OLIR would not constrain

FEDERAL R ESERVE BANK OF ST. LOUIS

short-run policymaking in undesirable ways, it would
nevertheless also help the market make inferences
about the likely timing and extent of tightening and
easing cycles, since, all else equal, the FOMC would
want the inflation rate to move ‘asymptotically’
toward the long-run desired level” (p. 167). Bernanke
also notes that the OLIR “would also serve as a
reminder to policymakers to keep one eye on the
long run at the same time that they are reacting to
current developments in the economy” (p. 167).
Noting that “sharing the OLIR with the public
would address the most important information
asymmetry in the system: namely, the public’s
imperfect knowledge of the FOMC’s objectives,”
Bernanke suggests that the OLIR be a “single number” and not a range (p. 168).
Finally, addressing the issue of the political
feasibility of IT for the United States, Bernanke suggests that “the change of the type I am proposing
would be acceptable to Congress as being within
the spirit of existing legislation.” To enhance that
prospect, Bernanke suggests
The FOMC might say to Congress: “We don’t
want long-run inflation to be too high,
because low inflation promotes growth and
productivity. On the other hand, inflation
shouldn’t be too low, because we want to
have all the room we need to respond to the
dangers that deflation poses for output and
employment. We pose the objective in terms
of inflation only because that is what the Fed
can control in the long run.” (p. 168)
Benanke’s colleague on the Board of Governors,
Don Kohn, undertakes a similar analysis, but arrives
at a different conclusion. Kohn agrees that “price
stability—or its approximation at very low inflation—
is the appropriate primary long-term objective of
monetary policy, and achieving this objective is the
way that policy can best contribute to the long-term
welfare of the country” (p. 179). Nevertheless, he
suggests that “adopting IT, even in its softer versions,
would be a slight shift along the continuum of constrained discretion in the direction of constraint,
and the benefits of such a shift are unlikely to outweigh its costs” (p. 179).
Kohn suggests that the adoption of an explicit
long-run inflation objective “would result in lessthan-optimal attention being paid to stabilizing the
economy and financial markets,” because “IT implies
putting higher priority on hitting a particular inflation objective over the intermediate run than the

Piger and Thornton

Federal Reserve has done” (p. 180). To illustrate
this point, he cites the “opportunistic disinflation
period” from 1983 to 1997, a period when “the
Federal Reserve was well aware that inflation was
running above levels consistent with price stability
but concentrated on keeping inflation from rising,
not on reducing it” (p. 180). He notes that the Fed’s
“broader focus was especially evident in the reaction
to the threat to financial stability in the fall of 1998
and in the very aggressive easing in early 2001”
(p. 180). Kohn notes that “such responses would in
theory be available under flexible IT,” but wonders
“what would happen in practice.”
As for the benefits of IT, Kohn notes that “the
bulk of studies show that interest rates and inflation
are no more predictable in IT economies,” and suggests that “the burden of proof should be on the
advocates of IT to show that it would improve economic performance in non-IT economies—by providing either greater cyclical stability or better resource
allocation” (p. 180).
Echoing concerns raised by Meyer, Friedman,
and others, Kohn notes that for flexible ITers “communication tends to be clear but not especially transparent”: “Those other goals are the tough messy
stuff that does not fit into the IT framework very
well” (p. 181). In apparent agreement with Ben
Friedman that “words matter,” Kohn notes “There
is also a risk that communication will drive policy,
and so those (other) goals end up with less-thanoptimal attention” (p. 181). In the end, in contrast
with Bernanke, Kohn concludes
IT is not a cure-all for communication
problems…it might not even help much in
the markets where it really counts, and that
if simplicity of communications drives policy,
IT might lead to inferior economic outcomes.
(p. 181)
As for the politics of IT, Kohn notes that
unlike most other central banks, which
operate in a parliamentary system, we do
not have a “government” to interact with…
If we moved toward setting a goal for ourselves, perhaps even if we just defined price
stability, we would need to consult carefully
with both houses of the Congress and the
Administration and would need to judge
what, short of legislation, constituted a veto
by any of the people with whom we were
consulting. This process would be subtle
and difficult—but absolutely essential to
J U LY / A U G U S T 2 0 0 4

11

Piger and Thornton

protect our independence and preserve our
democratic legitimacy. (p. 181, emphasis
added)
With respect to the idea of defining price stability—“publishing a number or reference range that
makes more concrete our long-term inflation objective, without making a commitment to achieve that
objective in any given time frame”—Kohn notes that
he is still trying to make up his mind “on the balance
of costs and benefits of taking this step” (p. 181).
Nevertheless, he expresses skepticism of the benefits
and suggests that the costs are associated with “any
tendency for this definition to morph into a target
that unnecessarily constrains actions” and notes that
“the pressure to elevate price stability over economic
stability…would be accentuated by the fact that the
latter goal would not have a numerical value.”
In the final analysis Kohn concludes that
those who propose changes from a good
system have a high burden of proof. The
marginal benefits from improving a good
regime, by definition, are not likely to be
high. And any change must deal with the
uncertainties created by the law of unintended consequences. I have yet to be convinced that for the United States inflation
targeting has jumped those hurdles. (p. 183)
Otmar Issing is also skeptical of the benefits of IT.
After a careful analysis of the major developments
in our understanding of the effectiveness of monetary policy since the 1960s, Issing notes that the
“consensus view” is that the management of monetary policy should be delegated to an “independent
central bank” and that policymakers should “treat
the natural rate of unemployment as a given, and
not try to push unemployment below its natural rate”
(p. 170). He also notes that
The awareness of the limitations of monetary policy was also coupled with a better
understanding of the possible costs of inflation and the recognition that a low-inflation
environment is a necessary precondition for
long-run growth and an efficient allocation
of resources. (p. 170)
While acknowledging that IT has been successful—particularly “for countries starting from high
levels of inflation”—he notes that other central banks
have shown that “success in the management of
monetary policy is not confined to inflation-targeting
12

J U LY / A U G U S T 2 0 0 4

REVIEW
central banks” (p. 171). This stems, in part, from
the adoption of the consensus view of the necessity
of maintaining a low-inflation environment. Defining
IT as a “monetary policy framework that accords
overriding importance to the maintenance of price
stability,” Issing suggests that IT “offers no practical
guidance for the conduct of monetary policy beyond
identifying the primary objective” and “from a scientific perspective…imposes few empirically testable
restrictions on the implementation of monetary
policy” (p. 171).
Consequently, Issing focuses his remarks on a
narrower definition of IT consistent with a “monetary policy framework based on the adoption of a
monetary policy rule in which forecasts of future
inflation play a central role, either in the form of the
so-called instrument rules or of target rules” (p. 171).
Issing argues a major drawback of instrument rules is
that inflation forecasts are not “sufficient statistics on
the state of the economy” necessary for the conduct
of monetary policy (p. 173). He then notes that
Even in an ideal world in which the models
producing the forecast are properly specified,
the policymakers are not interested in the
result of the forecast per se but instead aim
at a consistent economic picture—or, to put
it differently, they aim at identifying the relevant shocks underlying the forecasts and
how different types of disturbances to the
economy imply different kinds of policy
responses. The relation between forecasts
and underlying shocks is clearly one-to-one
in many simple stylized models used in the
monetary policy literature, but this relation
clearly breaks down once we depart from
that simple set-up. So, once again, forecasts
of a few macrovariables cannot be sufficient
as statistics to determine monetary policy
action. (p. 173)
Issing notes that target rules are immune from
this problem because they generate output and
inflation forecasts that are consistent with the minimization of the proper loss function. He notes, however, that models underlying most targeting rules
“neglect any role that might be played by the monetary aggregate or financial frictions in the determination of price developments” (p. 173). Moreover,
after suggesting that “model misspecification is
something that economists and econometricians
have some difficulty acknowledging,” Issing notes
that “most advocates of inflation targeting…ulti-

FEDERAL R ESERVE BANK OF ST. LOUIS

Piger and Thornton

mately rely on a view of the economy whose essence
can be captured by no more than three equations…
with monetary quantities playing no role” (p. 173).
While noting that some economists regard such
models as “internally consistent and elegant,” Issing
argues that because “it rests upon what can certainly
be regarded as extreme assumptions about the role
of money in the economy…a central bank can legitimately question the usefulness of [such] a model
for monetary policy-setting” (p. 174).
Furthermore, apparently following up on his
earlier observation that “the structure of the economy changes over time in a way that is difficult to
anticipate and perceive in real time…makes the
debate on the aims of monetary policy and its appropriate framework so difficult to settle” (p. 169),
Issing notes that “there are instances that standard
macroeconomic models, which, by definition, are
constructed to replicate normal conditions and
regularities in the economy, cannot capture and
incorporate” (p. 176). Issing notes that on such
occasions, “careful judgment” and “consideration
of non-standard indicators and different interpretations of the evidence become especially relevant”
(p. 176).
In the end, Issing concludes that “there is no
clear-cut evidence to suggest that generally, and
according to some well-specified criteria, one specific
framework should be preferred to all others.” He
argues that the success of the United States and the
euro area confirm “something that I have always
believed: there is no ‘single’ or ‘best’ way to conduct
monetary policy and that different approaches or
frameworks can lead to successful policies and/or
be better adapted to different institutional, economic, and social environments” (p. 177).
Over the past two decades, a number of central
banks have adopted explicit inflation targeting as a
framework for conducting monetary policy. Further,
there is an active debate, perhaps most notably in
the United States, regarding whether non-IT central
banks should adopt the practice. The conference
proceedings address the core issues surrounding
the potential benefits, drawbacks, and feasibility of
IT adoption, providing evidence from both a theoretical and empirical perspective. The authors, discussants, and panelists have provided a wealth of useful
information to inform the debate over IT.

J U LY / A U G U S T 2 0 0 4

13

REVIEW

14

J U LY / A U G U S T 2 0 0 4

Inflation Targeting and Optimal Monetary Policy
Michael Woodford

S

ince the early 1990s, an increasing number
of countries have adopted explicit inflation
targets as the defining principle that should
guide the conduct of monetary policy. This development is often credited with having brought about
substantial reductions in both the level and variability of inflation in the inflation-targeting countries,
and is sometimes argued to have improved the
stability of the real economy as well.1
Inflation-forecast targeting, as a systematic decision procedure for the conduct of monetary policy,
was developed at central banks like the Reserve Bank
of New Zealand, the Bank of Canada, the Bank of
England, and the Bank of Sweden on a trial-and-error
basis, with little guidance from the academic literature on monetary policy rules. But the growing popularity of inflation targeting has more recently led to
an active literature that seeks to assess the desirability of such an approach from the standpoint of theoretical monetary economics. This literature finds
that an optimal policy regime—one that could have
been designed on a priori grounds to achieve the
highest possible degree of social welfare—might
well be implemented through procedures that share
important features of the inflation-forecast targeting
that is currently practiced at central banks like those
just mentioned. At the same time, the normative
literature finds that one ought, in principle, to be
able to do better through appropriate refinement
of the practices developed at these banks.
Here I survey some of the most important conclusions of this literature. I shall begin by reviewing
some of the respects in which inflation targeting as
1

For surveys of early experiences with inflation targeting, see Leiderman
and Svensson (1995) and Bernanke et al. (1999). King (forthcoming)
offers an optimistic assessment of the improvements made in the
conduct of monetary policy in the United Kingdom under inflation
targeting. For a more skeptical review of the lessons that can be gleaned
from experience to date, see Ball and Sheridan (forthcoming).

currently implemented represents a step toward
what the theory of optimal monetary policy would
recommend. In the final section of the paper, I then
summarize some of the more important respects
in which an optimal policy regime would go beyond
current practice. Finally, as a concrete illustration
of some of the general remarks that have been made
about the form of an optimal policy rule, in an
appendix I briefly discuss the quantitative character
of optimal policy in the context of the small econometric model for the United States presented in
Giannoni and Woodford (forthcoming).

1. ADVANTAGES OF AN EXPLICIT
TARGET FOR MONETARY POLICY
Discussions of the desirability of inflation targeting for one country or another are often at cross
purposes because of differing implicit assumptions
about precisely what inflation targeting would mean.
It is thus perhaps useful to be clear from the outset
about what I regard to be the defining features of
the approach to the conduct of policy with which I
am concerned. Probably the most critical feature is
the existence of a publicly announced, quantitative
target that the central bank is committed to pursue,
the pursuit of which structures both policy deliberations within the central bank and communications
with the public. As should become clear from the
discussion below, it is more important in my view
that there should be an explicit target for policy than
that it should be (in any strict sense) an inflation
target. In my view, the most distinctive and most
important achievement of the inflation-targeting
central banks has not been the reorientation of the
goals of monetary policy toward a stronger emphasis
on controlling inflation—this has occurred, but it
has been a worldwide trend over the past two
decades, neither limited to nor even necessarily

Michael Woodford is the Harold H. Helm ’20 Professor of Economics and Banking at Princeton University. The author thanks Stephanie Schmitt-Grohé
and Lars Svensson for helpful comments and the National Science Foundation for research support through a grant to the National Bureau of
Economic Research.
Federal Reserve Bank of St. Louis Review, July/August 2004, 86(4), pp. 15-41.
© 2004, The Federal Reserve Bank of St. Louis.

J U LY / A U G U S T 2 0 0 4

15

Woodford

most associated with the innovators in inflation
targeting, and has hardly required a fundamental
change in the traditional concerns of central
bankers—but rather the development of an approach
to the conduct of policy that focuses on a clearly
defined target, that assigns an important role to
quantitative projections of the economy’s future
evolution in policy decisions, and that is committed
to a high degree of transparency as to the goals of
policy, the decisions that are made, and the principles
that guide those decisions.
It is useful to begin by discussing why it is desirable for a central bank to commit itself to an explicit
target as the goal of its policy. The proposal that
banks should do so runs contrary to a common
instinct of central bankers, according to which it is
wise to say as little as possible in advance about what
one may do in the future. Because central banking
is a complex task, the argument goes, any explicit
target or policy rule would prove to be a straightjacket, preventing the full exercise of the judgment
of central bankers on behalf of society when unanticipated circumstances arise, as they invariably do.
Furthermore, even if a formula could be developed
that would adequately describe what a good central
banker should do, announcing it publicly would only
invite second-guessing by the public and politicians
of policy decisions that are best left in the hands of
professionals. The best approach, then, is to delegate
the task to the best possible people, grant them full
discretion, and require as little public comment as
possible on the way they practice their arcane art.
But while it is true that central banking is complex, reasoning of this kind misses a fundamental
point about the kind of problem that a central bank
is called upon to solve. Central banking is not like
steering an oil tanker, or even guiding a spacecraft,
that follows a trajectory that depends on constantly
changing factors, but that does not depend on the
vehicle’s own expectations about where it is heading.
Because the key decisionmakers in an economy
are forward-looking, central banks affect the economy as much through their influence on expectations
as through any direct, mechanical effects of central
bank trading in the market for overnight cash. As a
consequence, there is good reason for a central bank
to commit itself to a systematic approach to policy
that not only provides an explicit framework for
decisionmaking within the bank, but that is also
used to explain the bank’s decisions to the public.
16

J U LY / A U G U S T 2 0 0 4

REVIEW

1.1 Central Banking as Management of
Expectations
One important advantage of commitment to an
appropriately chosen policy rule is that it facilitates
public understanding of policy. It is important for
the public to understand the central bank’s actions,
to the greatest extent possible, not only for reasons
of democratic legitimacy—though this is an excellent
reason itself, given that central bankers are granted
substantial autonomy in the execution of their task—
but also in order for monetary policy to be most
effective. For not only do expectations about policy
matter, but, at least under current conditions, very
little else matters. Few central banks of major industrial nations still make much use of credit controls
or other attempts to directly regulate the flow of
funds through financial markets and institutions.
Increases in the sophistication of the financial system
have made it more difficult for such controls to be
effective, and in any event the goal of improvement
of the efficiency of the sectoral allocation of resources
stressed above would hardly be served by such
controls, which (if successful) inevitably create
inefficient distortions in the relative cost of funds
to different parts of the economy.
Instead, central banks restrict themselves to
interventions that seek to control the overnight
interest rate in an interbank market for central bank
balances (for example, the federal funds rate in the
United States). But the current level of overnight
interest rates as such is of negligible importance
for economic decisionmaking; if a change in the
overnight rate were thought to imply only a change
in the cost of overnight borrowing for that one night,
then even a large change (say, a full percentage
point increase) would make little difference to
anyone’s spending decisions. The effectiveness of
changes in central bank targets for overnight rates
in affecting spending decisions (and hence ultimately pricing and employment decisions) is wholly
dependent upon the impact of such actions upon
other financial-market prices, such as longer-term
interest rates, equity prices, and exchange rates.
These are plausibly linked, through arbitrage relations, to the short-term interest rates most directly
affected by central-bank actions; but it is the
expected future path of short-term rates over coming months and even years that should matter for
the determination of these other asset prices, rather
than the current level of short-term rates by itself.

FEDERAL R ESERVE BANK OF ST. LOUIS

Thus the ability of central banks to influence
expenditure, and hence pricing, decisions is critically
dependent upon their ability to influence market
expectations regarding the future path of overnight
interest rates and not merely their current level.
Better information on the part of market participants
about central bank actions and intentions should
increase the degree to which central bank policy
decisions can actually affect these expectations, and
so increase the effectiveness of monetary stabilization policy. Insofar as the significance of current
developments for future policy are clear to the private
sector, markets can to a large extent “do the central
bank’s work for it,” in that the actual changes in
overnight rates required to achieve the desired
changes in incentives can be much more modest
when expected future rates move as well.2
The importance of being able to influence expectations about future policy through means other
than the announcement of a new operating target
for the overnight interest rate becomes especially
clear when the zero lower bound on nominal interest
rates prevents further interest-rate cuts, in an environment where aggregate nominal expenditure is
nonetheless too low. This is the situation that Japan
has faced for more than four years now, and recently
there has been considerable discussion in the United
States as well as to whether the Fed is not nearly
“out of ammunition” with which to fight a possible
threat of deflation. The key to avoiding deflation and
economic contraction under such circumstances,
as Eggertsson and Woodford (2003) show, is to be
able to credibly commit to looser monetary policy
in the future.3 This requires explicit discussion of
2

There is evidence that this is already happening, as a result of both
greater sophistication on the part of financial markets and greater
transparency on the part of central banks, the two developing in a sort
of symbiosis with one another. Blinder et al. (2001, p. 8) argue that,
from early 1996 through mid-1999, one could observe the U.S. bond
market moving in response to macroeconomic developments that
helped to stabilize the economy, despite relatively little change in the
level of the federal funds rate; furthermore, they suggest that this
reflected an improvement in the bond market’s ability to forecast Fed
actions before they occur. Statistical evidence of increased forecastability of Fed policy by the markets is provided by Lange, Sack, and
Whitesell (2001), who show that the ability of Treasury bill yields to
predict changes in the federal funds rate some months in advance
has increased since the late 1980s.

3

The basic point about the importance of commitment regarding future
policy was first made by Krugman (1998); Eggertsson and Woodford
(2003) present a fully dynamic analysis and characterize the optimal
policy commitment in an optimizing model with staggered pricesetting. The conclusion obtained by Auerbach and Obstfeld (2003) is
not fundamentally different: In their analysis, it is actually only the
expected money supply at the time of exit from the liquidity trap that

Woodford

the way in which policy will be conducted in the
future; furthermore, Eggertsson and Woodford show
that the kind of commitment that is needed can be
best expressed in terms of a commitment to a form
of price-level target, which the central bank is committed to eventually hitting, even if the zero bound
requires the target to be undershot for some period
of time. While one might alternatively imagine a
direct commitment regarding the length of time for
which interest rates will remain low, the optimal
continuation time will depend on how real conditions in the economy develop (that cannot yet be
perfectly foreseen); it is thus easier to explain the
kind of commitment that is actually appropriate by
explaining the target that will have to be met in order
for the zero interest-rate policy to be abandoned.
The existence of an explicit target for policy has
similar advantages under more ordinary circumstances as well. An obvious consequence of the
importance of managing expectations is that a
transparent central-bank decisionmaking process
is highly desirable. This has come to be widely
accepted by central bankers over the past decade.
(See Blinder et al., 2001, for a detailed and authoritative discussion.) But it is sometimes supposed that
the most crucial issues are ones such as the frequency of press releases or the promptness and
detail with which the minutes of policy deliberations
are published. Instead, from the perspective suggested here, what is important is not so much that
the central bank’s deliberations themselves be public,
as that the bank give clear signals about what the
public should expect it to do in the future. The public
needs to have as clear as possible an understanding
of the rule that the central bank follows in deciding
what it does. Inevitably, the best way to communicate
about this will be by offering the public an explanation of the decisions that have already been made;
the bank itself would probably not be able to describe
how it might act in all conceivable circumstances,
most of which will never arise.
The Inflation Reports of the leading inflationtargeting central banks provide good practical
examples of communication with the public about
the central bank’s policy commitments. These
reports do not pretend to give a blow-by-blow
account of the deliberations by which the central
bank reached the position that it has determined to
matters, so that open market operations while in the trap are effective
only to the extent that they are understood as implying a commitment
to a higher money supply after the zero bound ceases to bind.

J U LY / A U G U S T 2 0 0 4

17

Woodford

announce; but they do explain the analysis that
justifies the position that has been reached. This
analysis provides information about the bank’s
systematic approach to policy by illustrating its
application to the concrete circumstances that have
arisen since the last report; and it provides information about how conditions are likely to develop in
the future through explicit discussion of the bank’s
own projections. Because the analysis is made public,
it can be expected to shape future deliberations; the
bank knows that it should be expected to explain
why views expressed in the past are not later being
followed. Thus a commitment to transparency of
this sort helps to make policy more fully rule-based,
as well as increasing the public’s understanding of
the rule.
It might be argued that it should be enough for
a central bank to follow a systematic rule in its conduct of policy, without also needing to explain it to
the public. If one assumes rational expectations on
the part of the public, it would follow that the systematic pattern in the way that policy is conducted
should be correctly inferred from the bank’s observed
behavior. Yet while it would be unwise to choose a
policy whose success depends on its not being
understood by the public—which is the reason for
choosing a policy rule that is associated with a
desirable rational expectations equilibrium—it is
at the same time prudent not to rely too heavily on
the assumption that the public will understand policy
perfectly regardless of the efforts that are made to
explain it. Insofar as explanation of the policy rule
to the public does no harm under the assumption
of rational expectations, but improves outcomes
under the (more realistic) assumption that a correct
understanding of the central bank’s policy commitments does not occur automatically, then it is clearly
desirable for the central bank to explain the rule
that it follows.
The advantages of a public target when the
private sector must otherwise forecast future policy
by extrapolating from experience are shown in a
recent analysis by Orphanides and Williams (forthcoming). In the Orphanides-Williams model, private
agents forecast inflation using a linear regression
model, the coefficients of which are constantly
reestimated using the most recent observations of
inflation. The assumption of forecasting in this manner (on the basis of a finite time-window of historical
observations), rather than a postulate of rational
expectations, worsens the trade-off between inflation
variability and output-gap variability that is available
18

J U LY / A U G U S T 2 0 0 4

REVIEW
to the central bank. Allowing inflation variations in
response to “cost-push” shocks for the sake of outputgap stabilization is more costly than it would be
under rational expectations, because temporary
inflation fluctuations in response to the shocks can
be misinterpreted as indicating different inflation
objectives on the part of the central bank. Orphanides
and Williams then show that a credible commitment
to a long-run inflation target—so that private agents
do not need to estimate the long-run average rate of
inflation, but only the dynamics of transitory departures from it—allows substantially better stabilization
outcomes, though still not quite as good as if private
agents were to fully understand the equilibrium
dynamics implied by the central bank’s policy rule.
This provides a nice example of theoretical support
for the interpretation given by Mervyn King (forthcoming) and others of practical experience with
inflation targeting, which is that tighter anchoring
of the public’s inflation expectations has made possible greater stability of both real activity and inflation.

1.2 Avoiding the Pitfalls of Discretionary
Policy
There is also a further, somewhat subtler, reason
why explicit commitment to a target or policy rule
is desirable, given the forward-looking behavior of
the people in the economy that one seeks to stabilize.
It is not enough that a central bank have sound
objectives (reflecting a correct analysis of social
welfare); that it make policy in a systematic way,
using a correct model of the economy and a staff
that is well-trained in numerical optimization; and
that all this be explained thoroughly to the public.
A bank that approaches its problem as one of optimization under discretion—deciding afresh on the
best action in each decision cycle, with no commitment regarding future actions except that they will
be the ones that seem best in whatever circumstances
may arise—may obtain a substantially worse outcome, from the point of view of its own objectives,
than a bank that commits itself to follow a properly
chosen policy rule. As Kydland and Prescott (1977)
first showed, this can occur even when the central
bank has a correct quantitative model of the policy
trade-offs that it faces at each point in time, and the
private sector has correct expectations about the
way that policy will be conducted.
At first thought, discretionary optimization might
seem exactly what one would want an enlightened
central bank to do. All sorts of unexpected events

FEDERAL R ESERVE BANK OF ST. LOUIS

constantly occur that affect the determination of
inflation and real activity, and it is not hard to see
that, in general, the optimal level of interest rates
at any point in time should depend on precisely
what has occurred. It is plainly easiest, as a practical
matter, to arrange for such complex state-dependence
of policy by having the instrument setting at a given
point in time be determined only after the unexpected shocks have already been observed. Furthermore, it might seem that the dynamic programming
approach to the solution of intertemporal optimization problems provides justification for an approach
in which a planning problem is reduced to a series
of independent choices at each of a succession of
decision dates.
But standard dynamic programming methods
are valid only for the optimal control of a system
that evolves mechanically in response to the current
action of the controller. The problem of monetary
stabilization policy is of a different sort, in that the
consequences of the central bank’s actions depend
not only upon the sequence of instrument settings
up until the present time, but also upon privatesector expectations regarding future policy. In such
a case, sequential (discretionary) optimization leads
to a suboptimal outcome because, at each decision
point, prior expectations are taken as given, rather
than as something that can be affected by policy.
Nonetheless, the predictable character of the central
bank’s decisions, taken from this point of view, do
determine the (endogenous) expectations of the
private sector at earlier dates, under the hypothesis
of rational expectations: A commitment to behave
differently, that is made credible to the private sector,
could shape those expectations in a different way;
and because expectations matter for the determination of the variables that the central bank cares
about, in general, outcomes can be improved
through shrewd use of this opportunity.
The best-known example of a distortion created
by discretionary optimization is the “inflation bias”
analyzed by Kydland and Prescott (1977) and Barro
and Gordon (1983). In the presence of a short-run
“Phillips curve” trade-off between inflation and real
activity (given inflation expectations) and a target
level of real activity higher than the one associated
with an optimal inflation rate (in the case of inflation
expectations also consistent with that optimal rate),
these authors showed that discretionary optimization leads to a rate of inflation that is inefficiently
high on average, owing to neglect of the way that
pursuit of such a policy raises inflation expectations

Woodford

(causing an adverse shift of the short-run Phillips
curve). A commitment to an inflation target is one
obvious way of eliminating the temptation of suboptimal behavior of this particular kind.
However, many central bankers would argue
that they have absorbed the lesson of the KydlandPrescott and Barro-Gordon models and are able to
avoid systematically higher inflation than is desirable, without any need for advance commitments
regarding future policy. For example, they may view
themselves as using their discretion to minimize a
loss function that differs from the ones assumed
by Kydland and Prescott or Barro and Gordon in a
way that eliminates the high predicted average rate
of inflation in the Markov equilibrium associated
with discretionary policy.
In response to this, it is important to note that
the distortions resulting from discretionary optimization go beyond simple bias in the average levels
of inflation or other endogenous variables; this
approach to the conduct of policy generally results
in suboptimal responses to shocks as well. For example, various types of real disturbances can create
temporary fluctuations in what Wicksell called the
“natural rate of interest,” meaning that the level of
nominal interest rates required to stabilize both inflation and the output gap varies over time (Woodford,
2003, Chap. 4). However, the amplitude of the
adjustment of short-term interest rates can be more
moderate—and still have the desired size of effect on
spending and hence on both output and inflation—
if it is made more persistent, so that when interest
rates are increased, they will not be expected to
quickly return to their normal level, even if the real
disturbance that originally justified the adjustment
has dissipated. Because aggregate demand depends
upon expected future short rates as well as current
short rates, a more persistent increase of smaller
amplitude can have an equal affect on spending. If
one also cares about reducing the volatility of shortterm interest rates, a more inertial interest-rate policy
of this kind will be preferable; that is, the anticipation that the central bank will follow such a policy
leads to a preferable rational-expectations equilibrium (Woodford, 1999a; 2003, Chap. 7). But a central
bank that optimizes under discretion has no incentive to continue to maintain high interest rates
once the initial shock has dissipated; at this point,
prior demand has already responded to whatever
interest rate expectations were held then, and the
bank has no reason to take into account any effect
upon demand at an earlier date in setting its current
interest rate target.
J U LY / A U G U S T 2 0 0 4

19

REVIEW

Woodford

This distortion in the dynamic response of
interest rate policy to disturbances cannot be cured
by any adjustment of the way in which alternative
possible future paths for the economy are ranked
(assuming that the ranking depends only on the
future paths of inflation and other welfare-relevant
variables); instead, policy must be made historydependent, i.e., dependent upon past conditions even
when they are no longer relevant to the determination of the current and future evolution of the variables that the bank cares about. In general, no purely
forward-looking decision procedure—one that makes
the bank’s action at each decision point a function
solely of the set of possible paths for its target variables from that time onward—can bring about
optimal equilibrium responses to disturbances.
Discretionary optimization is an example of such a
procedure, and it continues to be when the bank’s
objective is modified, if the modified policy objective
still involves only the future values of the welfarerelevant variables. A commitment to use policy to
achieve a pre-specified target, instead, can solve this
problem if the target is defined in a way that incorporates the proper history-dependence.4
The advantages of an explicit target in solving
this kind of problem are especially clear in the case
of a binding zero lower bound on interest rates, as
discussed by Eggertsson and Woodford (2003). When
the natural rate of interest is temporarily negative,
the zero bound may prevent stabilization of inflation
and the output gap at their desirable long-run average
levels, as such an equilibrium would require a temporarily negative nominal interest rate. The key to
preventing an undesirably sharp deflation and economic contraction is to convince people that the price
level will eventually be raised, rather than being
stabilized at whatever level it may fall to in the period
during which the zero bound binds. A central bank
that is expected to optimize under discretion will not
be expected to subsequently undo the price decline
that occurs during the “liquidity trap” because the
absolute level of prices is not welfare-relevant; it will
therefore simply stabilize inflation and the output
gap once this again becomes possible, accepting
whatever level of prices happens to exist at that time.
A commitment in advance to the achievement of a
price level target—a target that is not allowed to shift
down even if actual prices undershoot it for many

quarters in a row, owing to the zero bound5—will
instead create expectations of the right sort. The
farther prices fall while the economy is in the “trap,”
the greater the expected future price increases will
be; and this automatic increase in expected inflation
will tend to prevent prices from falling very far, or
demand from contracting very much, in the first
place.
It is furthermore desirable not simply that a
central bank have a private intention of this sort,
but that it be publicly committed to such a target.
First, a public commitment is likely to make it easier
for the central bank’s policy deliberations to remain
focused on the right criterion—the criterion with the
property that systematic conformity to it leads to
an optimal equilibrium—rather than being tempted
to “let bygones be bygones.” And second, the benefits
associated with commitment to a history-dependent
policy depend entirely on this aspect of policy being
anticipated by the private sector; otherwise, it would
be rational to “let bygones be bygones.” There is no
point to a secret commitment to the future conduct
of policy in accordance with a history-dependent
rule while the private sector continues to believe
that the central bank will act in a purely forwardlooking fashion; thus the target should be explained
as clearly as possible to the public and shown to be
guiding the bank’s decisions.

1.3 Targeting Procedures as Policy Rules
It follows from the above discussion that there
are important advantages to a central bank’s commitment to conduct policy in accordance with a
rule that can be explained to the public in advance.
I turn now to the advantages of the particular type
of rule that is followed by the inflation-targeting
central banks. This is a rule under which the central
bank’s commitment is defined by a target for certain
variables at a certain distance in the future, together
with a commitment to organize deliberations about
policy actions around the question of whether the
contemplated actions are consistent with the target.
Much of the theoretical discussion of “rules
versus discretion” since the seminal contribution
of Kydland and Prescott (1977) has supposed that
the conduct of policy in accordance with a “rule”
would mean something rather different from this.
5

4

The kind of modified inflation target that leads to optimal responses
to the kinds of fluctuations in the natural rate of interest described
above is derived in Giannoni and Woodford (forthcoming, Section 1.3).

20

J U LY / A U G U S T 2 0 0 4

As Eggertsson and Woodford (2003) show, under an optimal policy
the price-level target would actually shift up in response to the target
misses during the period in which the zero bound is binding, and to a
greater extent the greater the target misses and the longer they persist.

FEDERAL R ESERVE BANK OF ST. LOUIS

On the one hand, an important branch of the literature on policy rules has emphasized the importance
of limiting central bank discretion, in the sense of
any scope for the exercise of judgment as to the
nature of current conditions. A rule is then considered, by definition, to be a prescription of a fairly
mechanical type, the dictates of which are unambiguous; it cannot pretend to allow optimal responses
to all of the different types of shocks that an economy may face, and indeed it is often asserted that
adherence to a rule means abandoning any concern
for the stabilization of real variables. Inflation forecast targeting as actually practiced is nowhere as
rigid a framework as this; in particular, projections
of the economy’s future evolution under alternative
possible actions play a central role in policy deliberations, and these projections, even when disciplined
by the use of a quantitative model, allow a rich range
of information about current conditions to be taken
into account in a way that could not be easily specified in advance by a computer program.
Alternatively, another branch of the literature
identifies “commitment” with a once-and-for-all
choice (at some initial date) of an optimal statecontingent plan for the central bank, which is implemented afterward by simply observing the state of
the world each period and executing the instrument
setting called for at that date and in that state. Under
the conception of rule-based policy in this literature,
the central bank may in principle pay attention to
disturbances of all sorts; but there is no role in a
specification of the policy commitment for any
mention of targets for variables other than the instrument of policy itself (i.e., for anything besides a
state-contingent operating target for the overnight
interest rate).
The type of rule actually followed, at least in
principle, by central banks like the Bank of England
is a policy rule of a different sort. Svensson (1999,
2003a) defines a targeting rule as a commitment to
adjust the bank’s policy instrument as necessary to
ensure that at each decision point the economy’s
future evolution is still projected to satisfy a certain
target criterion. For example, in the case of the Bank
of England, the target criterion is that CPI inflation
should be projected to equal 2 percent per annum
at a horizon eight quarters in the future.6 This is a
6

Before December 2003, the target criterion instead required that an
alternative measure of inflation, RPIX inflation, equal 2.5 percent eight
quarters in the future. For discussion of the role of this criterion in
the conduct of monetary policy in the United Kingdom, see Vickers
(1998) and Goodhart (2001).

Woodford

“higher-level” specification of a policy rule than the
kind generally considered in the two literatures just
referred to, since it leaves unspecified precisely what
policy actions will be required in any given circumstance to conform to the rule. Implementation of
the policy is only possible using a model of the
economy (likely to be supplemented, in practice,
by judgmental adjustments on the part of the monetary policy committee [MPC]), with which projections of the economy’s evolution under alternative
hypothetical policy decisions can be constructed.
Commitment to a decision procedure of this
kind has important advantages over both of the
other two conceptions of a monetary policy rule.7
Achievement of the advantages of policy commitment—in particular, avoidance of the inflationary
bias of discretionary policymaking—does not require
one to give up on stabilization policy. Not only may
policy adjust in response to disturbances, but it may
adjust differently to each of an uncountable number
of different types of disturbances, the nature of
which need not even be specifiable in advance.
This is also true, in principle, under the conception of a policy rule as a commitment to a prespecified state-contingent instrument path. But in
practice one cannot imagine computing such an
instrument rule in advance, and announcing one’s
commitment to it, unless one artificially assumes
that the number of different types of disturbances
that could occur is extremely limited. This is a highly
limiting assumption, given that in order to compute
in advance the optimal dynamic response to a given
shock, it is necessary not simply to specify which
equations of one’s model that it perturbs, but also
to give a detailed quantitative specification of the
dynamics of the shock—exactly how persistent it
is expected to be, how far in advance it can be predicted, and so on. Shocks of a given type—for example, variations in government spending owing to the
outbreak of war—that differ in the degree to which
they are unanticipated or the length of time for
which they are expected to last imply different
optimal adjustments of the policy instrument. Thus
they must be treated as different shocks in a complete
specification of the optimal state-contingent instrument rule.
Of course, one can specify a quantitative model
of the economy with a fairly small number of inde7

For further discussion, see Svensson (1999, 2003a), Svensson and
Woodford (forthcoming), Giannoni and Woodford (2002), and
Woodford (2003, Chap. 7).

J U LY / A U G U S T 2 0 0 4

21

Woodford

pendent shocks (no more than the number of
endogenous variables in the model) and estimate
a joint stochastic process for those shocks using
historical data. This method is often used, for example, in specifying the kind of stochastic model that
is used for “stochastic simulation” exercises evaluating alternative simple policy rules. And it may well
be possible to calculate a complete specification of
the optimal state-contingent instrument path for
such a model. But it would be highly unlikely for a
central bank to be willing to commit itself to follow a
rule simply because it has been shown to be optimal
in such an exercise.
For central bankers always have a great deal of
highly specific information about the kind of disturbances that have just occurred, which are always
somewhat different from those that have been faced
at other times. Hence even if it is understood that,
“typically,” disturbances to the level of military purchases have had a coefficient of serial correlation
of 0.9 at the quarterly frequency, there will often be
grounds to suppose that the conflict that is currently
looming is likely to be either more persistent or less
persistent than the “typical” one has been in the
past. And it is unlikely that central bankers will be
willing to commit themselves to stick rigidly to a
rule that is believed to lead to outcomes that would
be optimal in the case of “typical” disturbances, even
in a case in which they are aware of the economy
instead being subjected to “atypical” disturbances.
For a proposed policy rule to be of practical interest,
it must instead be believed that the rule is compatible
with optimal (or at least fairly good) outcomes for
the extremely large number of possible types of
disturbances. Yet if one were to try to write out the
optimal state-contingent instrument path, allowing separate terms for each of the possible (finely
grained) types of disturbances that might actually
be faced, such a description of optimal policy would
be completely unwieldy.
Giannoni and Woodford (2002) instead show
that if the central bank’s policy commitment is
described in terms of a relation among endogenous
variables that the bank is committed to bring about—
rather than in terms of a mapping from exogenous
states to the instrument setting—it is possible, in a
large class of policy problems, to find a rule that is
robustly optimal, in the sense that the same rule
(with given numerical coefficients) continues to be
optimal regardless of the assumed statistical properties of the (additive) disturbance terms in the model.
Indeed, the target criterion that the authors derive
22

J U LY / A U G U S T 2 0 0 4

REVIEW
characterizes optimal policy even if the disturbance
terms in the model structural equations are actually
composites of an extremely large (not necessarily
finite) number of different types of real disturbances.
This is possible because (as illustrated in the next
section) the optimal target criterion is derived from
certain first-order conditions that characterize an
optimal evolution of the economy, and these firstorder conditions do not involve the additive disturbance terms in the structural relations.
A rule of this kind represents a policy commitment that a central bank could reasonably make,
despite its awareness that it will constantly be receiving quite fine-grained information about current
conditions. For a belief that the target criterion
represents a sound basis for judging whether policy
is on track does not require the central bank to
believe that all shocks are alike, or even that all of
the possible types of disturbances to which it may
have to respond can all be listed in advance. At the
same time, a public commitment to the target criterion tells the public in advance what it should expect
with regard to the outcome to be achieved by policy.
This is actually what the public most needs to be
able to forecast well, and this is the aspect of the
public’s expectations that the central bank needs to
influence, to achieve the benefits that are available
in principle from policy commitment.

2. THE CASE FOR PRICE STABILITY
As noted above, the most important innovation
of the inflation-targeting central banks, in my view,
is the organization of policy deliberations around
the achievement of an explicit target, quite apart
from the type of target that happens to be chosen.
But another distinctive feature of inflation targeting,
of course, is that the target is for some measure of
inflation; while control of inflation has always been
an important concern of central bankers, inflation
targeting has given special, and sometimes exclusive,
emphasis to this goal, and debates about the desirability of inflation targeting are often primarily discussions of the desirability of such a strong emphasis
on inflation. Here I review what the theory of optimal
monetary policy has to say about this.
First of all, the modern (micro-founded) literature on the real effects of monetary policy provides
ample justification for the conventional wisdom of
central bankers—that it is better for inflation to be
both low and stable. It has been understood for some
time that expected inflation creates distortions by
increasing the opportunity cost of holding (non-

FEDERAL R ESERVE BANK OF ST. LOUIS

interest-earning) money, leading to the inefficient
use of real resources to economize on the use of
money in transactions; this was the basis for the
celebrated analysis of the optimal rate of inflation
(which actually turned out to be mild deflation) by
Friedman (1969).8 However, models that incorporate
some reason for prices to not adjust fully and instantaneously to changing market conditions—whether
these involve infrequent price changes or simply
slow updating of the information on which prices
are being set—imply that unanticipated variations
in the inflation rate create real distortions as well,
by causing prices that adjust at different times (or
that are being set on the basis of different information sets) to become misaligned with one another.
So price stability has important advantages, in
helping the market mechanism to work more effectively. Still, should this stabilization objective be given
priority over others, such as stabilization of real
economic activity or employment? I shall argue,
below, that it should not be an absolute priority; but
the recent literature on the welfare consequences
of alternative monetary policies finds that there is
less tension between inflation stabilization and
properly defined real stabilization objectives than
the traditional (non-welfare-theoretic) literature on
monetary stabilization policy has often suggested.
It is not a bad first approximation to say that the
goal of monetary policy should be price stability.

2.1 When Full Price Stability Is Optimal
Even when one grants that the economy is
subject to exogenous real disturbances of many
sorts—including various types of “supply shocks,”
i.e., disturbances that shift the “natural rate of output,” the level of output that would occur in an
equilibrium with fully flexible prices—it is possible
for the optimal monetary policy to be one that maintains completely stable prices in the face of these
disturbances and instead allows real activity to vary.
In particular, this is true in a wide variety of “sticky
price” or “sticky information” models (under varying
assumptions about how many price-setters revise
their prices or update their information in a given
interval of time), as discussed in Woodford (2003,
Chap. 6), where (i) the equilibrium fluctuations in
the real allocation of resources would be optimal if
only all prices were perfectly flexible and set on

Woodford

the basis of fully up-to-date information and (ii)
there are only aggregate shocks, so that in a flexibleprice equilibrium all goods would have the same
price. These hypotheses allow for the existence of
a wide range of types of real aggregate disturbances
that should affect the natural rate of output—for
example, exogenous variation in technology, in
preferences regarding labor supply or impatience
to consume, or in government purchases—though
it does not allow for certain kinds of “supply shocks,”
such as variations in the degree of market power in
labor or product markets, or variations in tax rates.
The basic intuition is fairly simple.9 The deadweight losses due to relative price distortions can be
completely eliminated, in principle, by stabilizing
the aggregate price level. For the aggregate price
level is stabilized by creating an environment in
which suppliers who choose a new price (under
full information) have no desire at any time to set a
price different from the average of existing prices.
Then (because the average price level never changes),
the price desired by any supplier that reconsiders
its price is always the same, regardless of the number
of future periods for which the price is expected to
remain fixed and regardless of how incomplete the
supplier’s information may be about current market
conditions. All new prices are then always chosen
to equal the average of existing prices, and as a result
the average price never changes. And all goods
prices must eventually equal that same, constant
value, so that inefficient relative-price dispersion
due to price stickiness or information imperfections
will not exist.
Furthermore, in such an environment, the
equilibrium real allocation of resources will be the
same as if all prices were fully flexible and set under
perfect information. For by hypothesis, in that case
suppliers would also all choose a common price
equal to the current price index. Since they are able
to charge this price at all times despite the infrequency of their reconsideration of their prices or
the limitations of the information that they can use
in adjusting prices, neither the stickiness of prices
nor that of information has any effect on equilibrium
behavior. Since, by hypothesis, the equilibrium
allocation of resources would be optimal under full
information and full price flexibility, it is optimal
under the monetary policy that fully stabilizes prices.
9

8

Friedman’s argument remains correct in the case of a wide range of
different ways of modeling the source of the demand for money; see,
e.g., Woodford (1990).

It is presented in the case of a model of staggered pricing by Goodfriend
and King (1997). The fact that a similar conclusion is obtained in the
case of “sticky information” is illustrated by the analysis of Ball,
Mankiw, and Reis (2003).

J U LY / A U G U S T 2 0 0 4

23

Woodford

2.2 Qualifications
The conditions under which full price stability
can be shown to be an optimal policy are in some
respects quite general; for example, the conclusion
does not depend on fine details of how many prices
are set a particular time in advance or left unchanged
for a particular length of time. Nonetheless, the conditions assumed above are quite special in other
respects—at least as an exact description of reality—
and it is likely that some degree of deviation from
full price stability is warranted in practice. Some of
the more obvious reasons for this are sketched here.
First of all, complete price stability may not be
feasible. In the argument sketched above, I have
supposed that it is possible to use monetary policy
to maintain an environment in which a supplier with
flexible prices and full information would never wish
to change its price. Often there will exist a statecontingent path for short-term nominal interest rates
consistent with such an equilibrium; it is shown in
Woodford (2003, Chap. 4) that this requires that the
interest rate track the Wicksellian natural rate of
interest—the real rate of return that would prevail
in an equilibrium with flexible prices and full information—which varies in response to real disturbances. However, it is possible that at some times
(as a result of exogenous real disturbances of a
particular sort) the natural rate of interest is temporarily negative; if so, there cannot be an equilibrium in which the nominal rate of interest is equal
at all times to the natural rate, and hence no equilibrium in which inflation is zero at all times. As a result,
a policy will have to be pursued that involves less
volatility of the short nominal interest rate in
response to shocks, and some amount of price stability will have to be sacrificed for the sake of this.
Varying nominal interest rates as much as the
natural rate of interest varies may also be desirable
as a result of the “shoe-leather costs” involved in
economizing on money balances. As argued by
Friedman (1969), the size of these distortions is
measured by the level of nominal interest rates, and
they are eliminated only if nominal interest rates
are zero at all times. Taking account of these distortions—from which we have abstracted thus far10—
10

The hypothesis above that the equilibrium allocation of resources
was efficient under flexible prices required, among other things, that
transactions frictions of this kind be abstracted from. The economies
referred to in the previous section are “cashless,” or at least nearcashless economies, in which transactions frictions are unimportant.
See Woodford (2003) for details.

24

J U LY / A U G U S T 2 0 0 4

REVIEW
provides another reason for the equilibrium with
complete price stability, even if feasible, not to be
fully efficient; for as Friedman argues, a zero nominal
interest rate will typically require expected deflation
at a rate of at least a few percent per year.
And taking account of these distortions affects
more than the optimal average rate of inflation. As
with distorting taxes, it is plausible that the deadweight loss resulting from a positive opportunity
cost of holding money is a convex function of the
relative price distortion, so that temporary increases
in nominal interest rates are more costly than temporary decreases of the same size are beneficial. In
short, monetary frictions provide a further reason
to reduce the variability of nominal interest rates,
even taking as given their average level. (At the same
time, reducing their average level will require less
variable rates, because of the zero floor.) Insofar as
these costs are important, they too will justify a
departure from complete price stability, in the face
of any real disturbances that cause fluctuations in
the natural rate of interest, to allow greater stability
of nominal interest rates.
Yet while both of the factors just mentioned
justify some departure from complete price stability,
it is not clear that the volatility of inflation should
be very great under an optimal policy, even when
such factors are taken account of. For example,
Rotemberg and Woodford (1997) characterize optimal policy for an estimated model of the U.S. economy, when a constraint that the mean federal funds
rate must remain at least a certain number of standard deviations (greater than two) above zero is
imposed as a substitute for the zero bound (that still
allows a linear characterization of optimal policy).
Even though the real disturbance processes in their
model imply greater volatility of the natural rate of
interest than many would assume (a standard deviation between 3 and 4 percentage points), they find
that optimal policy involves an average rate of inflation only slightly greater than zero (11 basis points!)
and not much variability of inflation (a standard
deviation only 40 percent as large as the actual variability of U.S. inflation over their post-1980 sample period). Interest rates are smoothed considerably
in the optimal policy, relative to what would be
required to fully stabilize inflation, but this does not
require too much variation in inflation, as their estimated model implies a variance trade-off that is quite
flat near the extreme of full inflation stabilization.

FEDERAL R ESERVE BANK OF ST. LOUIS

For the same reason, taking account of the distortions created by high nominal interest rates in a
model with transactions frictions justifies only a
relatively modest degree of inflation variation for
the sake of greater stability of nominal interest rates,
at least if the transactions frictions are calibrated at
an empirically realistic magnitude. Woodford (2003,
Chap. 6) finds that when transactions frictions are
calibrated to match facts about U.S. money demand,
the penalty on nominal interest-rate variations that
can be justified on welfare-theoretic grounds is a
good bit smaller than the one assumed in Rotemberg
and Woodford (1997). Hence in the case that the
available trade-off between interest-rate variability
and inflation variability is the one estimated by
Rotemberg and Woodford, the degree of inflation
variability that could be justified on this ground
would be even smaller than in their paper.
Even apart from these grounds for concern with
interest-rate volatility, the class of models for which
full price stability is optimal is a special one in several
respects. One obvious restrictive assumption in the
argument sketched above is that there are assumed
to be no shocks that would require the relative prices
of any goods to vary over time in an efficient equilibrium (i.e., the shadow prices that would decentralize an optimal allocation of resources involve no
variation in relative prices). If, instead, an efficient
allocation of resources requires relative price
changes, due to asymmetries in the way that different
sticky-price commodities are affected by shocks,
then full stabilization of a symmetric index of prices
is not generally optimal, as shown by Aoki (2001)
and Benigno (forthcoming) in the context of twosector models with asymmetric disturbances.
Nonetheless, it may still be possible to define
an asymmetric price index, with the property that
stabilization of this index is optimal, at least a good
approximation to optimal policy, as these authors
show.11 If the model is symmetric except for the
frequency of adjustment of different types of prices,
then the optimal price index to stabilize puts more
weight on the prices of the goods with “stickier”
prices; this provides a theoretical justification for
targeting an appropriately constructed measure of
11

Benigno (forthcoming) applies this idea to an analysis of optimal stabilization objectives for a monetary union in which different regions are
affected asymmetrically by real disturbances. In this application, the
optimal inflation target for the monetary union does not necessarily put
weights on the national inflation rates that are proportional to the shares
of those country’s products in the union-wide consumption basket.

Woodford

“core” inflation, rather than a standard consumer
price index. But as long as the price index to be
stabilized is appropriately chosen, complete stabilization of a price index is found (in calibrated
examples) to be nearly optimal.
Similarly, the analysis sketched above assumed
flexibility of wages. While this is a familiar assumption in sticky-price models used for pedagogical
purposes, many empirical models imply that wages
are as sticky as prices, and possibly more so.12 But
real disturbances almost inevitably require real
wage adjustments in order for an efficient allocation
of resources to be decentralized. And if both wages
and prices are sticky, it will then not be possible to
achieve all of the relative prices associated with efficiency simply by stabilizing the price level—specifically, the real wage will frequently be misaligned,
as will be the relative wages of different types of
labor if these are not set in perfect synchronization.
In such circumstances, complete price stability
may not be a good approximation at all to the optimal policy, as Erceg, Henderson, and Levin (2000)
show. Nonetheless, one can show once again that
stabilization of an appropriately weighted average
of prices and wages may still be a good approximation to optimal policy; it is even fully optimal in
special cases (Woodford, 2003, Chap. 6). Thus concerns of this kind are not so much reasons not to
pursue price stability as they are reasons why care
in the choice of the index of prices (including wages)
that one seeks to stabilize may be important.
Finally, even when wages are flexible (or there
are efficient labor contracts) and all disturbances
have symmetric effects on all sectors of the economy, the flexible-price equilibrium level of output
need not be welfare-maximizing. Both market
power and the existence of distorting taxes imply
that in reality, the equilibrium level of economic
activity is likely to be too low on average.13 When
this is true, not only is the flexible-price equilibrium level of output different from the (first-best)
optimal level, but except in special cases, real disturbances will not shift these two quantities to
12

See, e.g., Amato and Laubach (2003), Christiano, Eichenbaum, and
Evans (2001), Altig et al. (2002), Smets and Wouters (2002a,b), and
Giannoni and Woodford (forthcoming).

13

This does not occur in the model of Rotemberg and Woodford (1997),
owing to the assumed presence of an output subsidy that offsets the
consequences of the market power of the monopolistically competitive
suppliers of differentiated goods.

J U LY / A U G U S T 2 0 0 4

25

Woodford

quite the same extent (in percentage terms).14 This
means that the gap between the level of output
associated with a policy that maintains stable prices
(which is the same as the flexible-price equilibrium
output, as explained above) and the optimal level
of output will be time-varying. If we write the
aggregate-supply relation as a relation between
inflation and the welfare-relevant “output gap”
(i.e., the gap between the actual and efficient levels
of output), an additional exogenous “cost-push” term
appears. As a consequence, it will not be possible
to simultaneously stabilize inflation and the welfarerelevant output gap.15
Yet even so, the degree of variability of inflation
under an optimal policy may be quite modest. This
is because the relative weight that should be placed
on the goal of output-gap stabilization, relative to
the weight on inflation stabilization, may not be
large. (This is illustrated in the welfare-based loss
function for the model of Giannoni and Woodford,
forthcoming, presented in the appendix.) There is
a straightforward reason for this. In a variety of optimizing models with sticky prices, it is shown in
Woodford (2003, Chap. 6) that the loss function
that corresponds to a quadratic approximation to
expected utility involves a relative weight on outputgap stabilization that is proportional to the coefficient
on the output gap in the short-run aggregate-supply
relation. This means that the same underlying
microeconomic factors that lead to a relatively flat
aggregate-supply relation—and thus imply that
fluctuations in nominal aggregate demand have
large effects on output relative to their effects on
prices—also imply that the welfare losses associated
with fluctuations in the level of aggregate real activity
are small relative to the welfare losses that result
from the misalignment of prices that are not adjusted
with perfect synchronization when inflation varies.
14

King and Wolman (1999) and Khan, King, and Wolman (2002) analyze
a model in which it is optimal to fully stabilize prices in response to
technology shocks, despite the existence of an inefficiently low steadystate level of output. This result, however, depends on the assumption
of special isoelastic functional forms for both preferences and technology, and also on the assumption of zero steady-state government
purchases; deviations from any of these assumptions will result in full
price stability no longer being optimal. Also, even under the assumed
specification, other types of real disturbances imply that it will not be
optimal to fully stabilize inflation, as Khan, King, and Wolman show.
See Woodford (2003, Chap. 6) for further discussion.

15

Even when the average level of output is efficient, the flexible-price
level of output and the efficient level may be differently affected by
certain kinds of real disturbances. As noted above, these include variations in market power or in the level of tax distortions.

26

J U LY / A U G U S T 2 0 0 4

REVIEW
It follows that, while the welfare-theoretic loss
functions derived for the estimated models of
Rotemberg and Woodford (1997) and Giannoni and
Woodford (forthcoming) involve stabilization goals
other than inflation stabilization, by far the largest
coefficients are those on the inflation stabilization
goal. Given this, optimal policy will still be focused
to an important degree on inflation stabilization.
While the considerations sketched in this section
give one ample reason to consider the consequences
of monetary policy for the evolution of variables
other than inflation, it will nonetheless make sense
to think of the optimal policy rule as a “flexible
inflation targeting rule.”

3. IMPROVING THE PRACTICE OF
INFLATION-FORECAST TARGETING
I turn now to some ways in which an optimal
forecast-targeting procedure for the conduct of
monetary policy, from the perspective of the theoretical literature summarized above, would differ from
inflation-forecast targeting as it is currently practiced
by the central banks that have led the way in developing this approach to monetary policy. Of course,
the precise details of an optimal procedure depend
on the details of one’s model of the monetary transmission mechanism, and it can hardly be argued
that there is yet a consensus about the correct model
to use for one country, let alone a model that can
be claimed to apply equally to all countries. Nonetheless, it seems that one can draw at least a few
broad lessons about the character of optimal policy
rules from the analyses that have been undertaken
thus far, and that these differ enough from current
practice to allow some suggestions for improvement.

3.1 The Target Criterion Should Involve
More Than Inflation
The official target criterion of the Bank of
England—ensuring that projected CPI inflation eight
quarters in the future should always equal 2 percent
per annum—refers only to the projected future
value of a particular measure of U.K. inflation. While
other inflation-targeting central banks are often less
explicit about the precise way in which current
policy decisions are supposed to be determined by
their inflation targets, it is very generally the case
that there is an explicit target only for (some measure of) inflation and no commitment to take into
account the projected paths of any other variables.
Hence debates about the desirability of inflation

FEDERAL R ESERVE BANK OF ST. LOUIS

targeting in countries such as the United States often
assume that such an approach to policy would mean
a sole concern with inflation stabilization.
An optimal policy, instead, will not involve complete stabilization of inflation except under fairly
special circumstances, as discussed in the previous
section. In general, an optimal policy will involve
some degree of temporary variation in the inflation
rate in response to real disturbances, for the sake
of greater achievement of other stabilization objectives. The degree to which this matters in practice
will depend on the quantitative specification of one’s
model of the economy; but an identification of
inflation targeting with what Svensson (1999) calls
“strict inflation targeting” makes it too easy for opponents of inflation targeting to argue that it would
prevent the central bank from responding in appropriate ways to changing economic conditions.
It is sometimes argued that a coherent monetary
policy requires “a single objective,” so that stabilization objectives in addition to inflation stabilization
should play no role in the conduct of monetary
policy, despite the admitted desirability of these
ends.16 It is true that a simultaneous commitment
to stabilize two different variables using a single
policy instrument will, in general, represent a promise that cannot possibly be fulfilled. But a commitment to a single target criterion, on the basis of which
the instrument of policy is to be adjusted, does not
require that this criterion involve only a single
variable. The target criterion may well be a linear
combination of projections for several different variables (just as it may also involve inflation projections
at more than one horizon). In general, an optimal
target criterion will be of this form. For example, in
the case of the Giannoni-Woodford model of the
U.S. monetary transmission mechanism discussed
in the appendix, the optimal target criterion involves
not only projected inflation, but also real-wage and
output-gap projections. A lower projection of real
wage growth or of the (welfare-relevant) output gap
will justify acceptance of a higher projected inflation
rate. Nonetheless, there is a single well-defined
measure at each point in time of whether policy
remains on track.
16

A related view asserts that other goals may be introduced only to the
extent that they do not interfere with achievement of the inflation
target. However, absolute priority of the inflation target would not
seem to leave any room for stabilization of output or other variables,
unless the inflation target is not understood to require stabilization
of inflation to the greatest extent possible. Such formulations are
thus hopelessly vague about what the policy commitment actually
promises.

Woodford

It is not obvious, of course, that actual inflationtargeting central banks do not take into account
other stabilization objectives in their policy decisions,
despite their use of an official rhetoric that suggests
a strict inflation target. Commentators such as
Bernanke et al. (1999) and Svensson (1999) argue
that all actual inflation-targeting central banks are
“flexible inflation targeters” that trade off inflation
stabilization against other stabilization objectives.
Furthermore, it is often argued that a particular
advantage of inflation-forecast targeting as a policy
rule is precisely that it allows monetary policy to be
used to reduce the short-run effects of disturbances
on real variables (such as the output gap), while
retaining firmly anchored medium-term inflation
expectations, and hence reducing the degree of inflation variability that is required to achieve a given
degree of stability of the real variables.
I do not doubt that actual inflation-targeting
central banks do take some account of real objectives. For example, the introductory summary of the
Bank of England’s Inflation Report always presents
a chart of the Bank’s current real GDP projection as
well as its inflation projection—and the GDP projection is always discussed first, even if it is solely the
inflation projection that is cited as showing that
policy is on track. But it would be desirable for central
banks to commit themselves to the pursuit of explicit
target criteria that involve real variables as well as
inflation. For one thing, if the criteria on which
policy is actually based include projections for other
variables, it would increase transparency, facilitating
the public’s ability to correctly anticipate future
policy, to explain policy in this way. In addition,
greater frankness about this aspect of banks’ policy
commitments would help to dispel some of the
resistance to the adoption of inflation targeting in
countries like the United States. In particular, it would
show that adoption of a targeting framework by
the Federal Reserve need not imply any departure
from the Fed’s current legal mandate—which
requires it to pursue full employment as well as
price stability—and hence need not wait for
Congressional authorization.17

3.2 A “Medium Term” Target Is Not
Enough
Many would argue that the reason that inflationtargeting central banks have only an unqualified,
17

On the issue of whether the adoption of inflation targeting in the
United States would require new legislative authority, see also
Goodfriend (forthcoming).

J U LY / A U G U S T 2 0 0 4

27

REVIEW

Woodford

time-invariant target for inflation—rather than a
target criterion that takes account of output projections, or other variables, as well—is that the inflation
target represents only a “medium-term” goal that
leaves unspecified the precise transition path by
which the medium-term goal is to be reached. (This
is explicit in the case of the Bank of England’s official
target criterion. Only the rate of inflation eight quarters in the future must equal the time-invariant target
rate; nearer-term inflation projections are allowed to
vary.) The appropriate medium-term inflation target
can be stated in an unqualified, time-invariant form,
it is argued, because there is no substantial trade-off
between the inflation rate and real variables this far
in the future. Other stabilization goals are instead
appropriately taken into account in choosing among
the possible nearer-term transition paths that are
consistent with the medium-term target.
In fact, the sort of optimal target criteria that
can be derived using the method of Giannoni and
Woodford (2002) involve much nearer-term projections than those that are officially targeted by the
Bank of England or other inflation-targeting central
banks. For example, while the targeting criteria discussed in the appendix involve weighted averages
of projections for many different future quarters, it
is the projection for one or two quarters in the future
that receives the greatest weight. Thus the optimal
target criteria do not merely describe the state that
one wishes to reattain once the effects of recent
disturbances have worked themselves out; they
also characterize the optimal transition dynamics
following a disturbance.
A simple example may be useful in clarifying
this. Suppose that the prices of individual goods are
re-optimized at random intervals as proposed by
Calvo (1983), but that all prices are fixed a quarter
in advance, so that even those new prices that are
chosen in quarter t take effect only beginning in
quarter t+1. Suppose furthermore that, between the
occasions on which the optimality of a given price is
reconsidered, it is automatically indexed to an aggregate price index (but, again, the aggregate price index
of the quarter before the one in which the price will
apply), as proposed by Christiano, Eichenbaum, and
Evans (2001). In a simple model with fixed capital
and no labor-market frictions, this results in an
aggregate-supply relation of the form18
18

This is essentially the form of aggregate-supply relation proposed by
Fuhrer and Moore (1995), and a simplified version of the aggregatesupply blocks of the empirical optimizing models of Christiano,
Eichenbaum, and Evans (2001), Altig et al. (2002), Smets and Wouters
(2002a, 2002b), and Giannoni and Woodford (forthcoming).

28

J U LY / A U G U S T 2 0 0 4

(3.1) π t − π t −1 = κ Et −1xt + β Et −1(π t +1 − π t ) + ut −1,
where πt is the quarter-t inflation rate, xt is the
(welfare-relevant) output gap, ut –1 is an exogenous
(mean-zero) random disturbance at date t–1, κ is a
positive coefficient, and β is the discount factor of
the representative household. Exogenous fluctuations in the “cost-push” term ut –1 as a result of various real disturbances then create a tension between
the goals of inflation stabilization and output-gap
stabilization.
Under the microeconomic foundations proposed
for the aggregate-supply relation above, the appropriate welfare-theoretic stabilization objective corresponds to minimization of a loss function of the
form19
`

(3.2)

[

]

Et 0 ∑ β t −t 0 (π t − π t −1 ) + λ ( xt − x *) ,
t = t0

2

2

where both the optimal output gap x* (positive in
the empirically realistic case) and the positive relative
weight λ depend on model parameters. Because it
is assumed that prices are automatically indexed to
a lagged aggregate price index, inflation creates
distortions in the model only to the extent that the
aggregate inflation rate differs from that in the previous quarter; hence policy should aim to stabilize
the rate of change of inflation, rather than its absolute
level.20 As we shall see, however, this does not mean
that it is not desirable for the central bank to commit
to a fixed long-run inflation target.
Let us consider now the problem of conducting
policy from some date t0 onward so as to minimize
(3.2), subject to a constraint
(3.3)

π t0 +1 = π t0 .

This last constraint prevents the policy authority
from choosing a policy at date t0 that fails to internalize the effects of policy at t0 (insofar as it could have
been forecasted in the previous quarter) on the
inflation-output trade-off faced in quarter t0 –1.
Choosing a policy commitment from date t0 onward
in the absence of any such constraint would result
in selection of a policy that is not time-consistent,
19

For details of the derivation, see Woodford (2003, Chap. 6).

20

The conclusion that the absolute level of inflation has no consequences
for welfare is extremely special to the simple case considered here,
and surely not realistic, as discussed in Woodford (2003, Chap. 7). For
similar analyses of the form of optimal target criteria when there is
no indexation, or only partial indexation, see Svensson and Woodford
(forthcoming) and Giannoni and Woodford (forthcoming).

FEDERAL R ESERVE BANK OF ST. LOUIS

for one would commit to a policy at all later dates
that took account of these effects. If the constraint
π– t0 is chosen (as a function of the state of the world
in quarter t0 ) in a “self-consistent” way, the optimization problem just posed can be solved by a timeinvariant policy rule. Furthermore, if one reconsiders
the desirability of following the policy rule at any
later date, then (assuming that one’s model of the
economy and policy objectives have not changed
in the meantime) one would continue to find that
the same time-invariant policy rule would continue
to solve the corresponding constrained optimization
problem looking forward from the later date.21
Finally, let us suppose that the component of
aggregate real expenditure that is sensitive to interest
rates is also determined a quarter in advance, so that
the output gap xt cannot be affected by monetary
policy decisions later than quarter t–1.22 It follows
that monetary policy can affect only the evolution
of inflation and the component of the output gap
that is forecastable a quarter in advance, and that the
possible stochastic paths for these variables that can
be achieved by any monetary policy are the set of
processes consistent with relation (3.1) for t ≥ t0+1.
The first-order conditions for the optimization
problem just stated are then of the form
(3.4)

π t +1 − π t + ϕ t − ϕ t −1 = 0

(3.5)

λ ( Et xt +1 − x *) − κϕ t = 0

for each t ≥ t0, where ϕ t–1 is the Lagrange multiplier
associated with constraint (3.1) for each t>t0, and
ϕ t0 –1 is a multiplier associated with the constraint
(3.3). These conditions, together with the constraints,
determine the optimal state-contingent evolution
of inflation and the forecastable component of the
output gap; the unforecastable component of the
output gap, of course, is exogenously given.
How should monetary policy be conducted to
ensure that this desired state-contingent evolution
of inflation and output is realized? Applying the
method of Giannoni and Woodford (2002), one can
21

22

See Woodford (2003, Chap. 7) for further discussion. A policy that
solves a problem of this form is “optimal from a timeless perspective,”
as discussed in Woodford (1999b).
For models of aggregate demand with this property, see Woodford
(2003, Chap. 5). This kind of predetermination of interest-sensitive
aggregate expenditure is a feature of many empirical optimizing
models, such as Rotemberg and Woodford (1997), Amato and Laubach
(2003), Christiano, Eichenbaum, and Evans (2001), Altig et al. (2002),
and Giannoni and Woodford (forthcoming), along with many ad hoc
macroeconometric models.

Woodford

eliminate the Lagrange multiplier from equations
(3.4) and (3.5), and show that one must have
(3.6)

(π t +1 − π t ) + φ ( Et xt +1 − Et −1xt ) = 0

for each t ≥ t0, where φ ; λ /κ>0. This in turn implies
that
(3.7)

π t +1 + φ Et xt +1 = π *

for each t ≥ t0, where π* is a constant, the value of
which will depend on the initial constraint π– t0. For
any value of π*, there exists a self-consistent specification of the initial constant under which optimal
policy satisfies (3.7) for all t ≥ t0. Thus the optimal
long-run inflation target π* is not determined within
this model.23
Optimal policy, then, must arrange that (3.7)
holds at each date, or equivalently, that
(3.8)

Et [π t +1 + φ xt +1 ] = π *.

(This alternative form emphasizes that the terms in
the target criterion can be affected only by monetary
policy decisions in quarter t or earlier.) Conversely,
one can show that if policy ensures that (3.8) is satisfied at each date t ≥ t0, the unique nonexplosive
rational-expectations equilibrium consistent with
the policy commitment solves the optimization
problem stated above. Hence (3.8) is an optimal
target criterion for the central bank’s policy decision
in quarter t.
In the model sketched above, the central bank
cannot expect to affect whether (3.8) holds in quarter
t through adjustment of the interest rate it in that
quarter, for the predetermination of the interestsensitive component of expenditure implies that
unforecastable interest-rate changes have no effect
on aggregate demand. The central bank’s period-t
policy decision should then be a commitment it+1,t
regarding its operating target for the interest rate in
quarter t+1. The value of it+1,t should be chosen so
as to lead the central bank to project that (3.8) is
satisfied, conditional on the state of the economy
in quarter t.24 The expectation that it+1,t will be
chosen in this way in each quarter t ≥ t0, and that the
23

The addition of even small frictions can break the indeterminacy of
the optimal long-run inflation target, as discussed in Woodford (2003,
Chap. 7). In practice, one can be certain that the optimal long-run
inflation target is not far from zero; it could even be slightly below zero.

24

Note that in the model sketched here, both Etπ t+1 and Et xt+1 should
be affected by the bank’s choice of it+1,t , assuming that the bank’s
announcement of its target for the following quarter is credible to the
private sector.

J U LY / A U G U S T 2 0 0 4

29

REVIEW

Woodford

central bank will then act to ensure that it+1=it+1,t
in the following quarter, will then imply the desired
state-contingent evolution of inflation and output.
The proposed policy rule involves a constant
long-run inflation target, since satisfaction of (3.8)
each quarter implies that one must have
lim Et π T = π *
(3.9)
T →`

at all times t. And it would surely be desirable for
the central bank to emphasize to the public its
commitment to a policy that implies (3.9), as this
should help to anchor long-run inflation expectations (which would never be allowed to vary in an
optimal equilibrium). Nonetheless, a commitment
to ensure that (3.9) is satisfied at all times is not
sufficient for optimality; many different sorts of
transitory responses to disturbances would be
equally consistent with it.
Nor is it clear what a central bank is committing
itself to do if it pledges to ensure that (3.9) is satisfied
at all times. Condition (3.9) does not place any restrictions on the behavior of interest rates over any finite
horizon; hence it is not clear what one would be
able to monitor about a central bank’s decisions in
order to verify that it is indeed acting in conformity
with its supposed commitment. Condition (3.8),
instead—together with the expectation that (3.8) will
also hold at all future dates—does imply a particular
rational-expectations equilibrium value for Et it+1
and so one could monitor, at least in principle,
whether it+1,t is chosen in accordance with it.
The same is true in the case of a “medium term”
target that refers to a specific future date. Condition
(3.8) implies that
(3.10)

λ


Et π t + k + xt + k  = π *
κ



must also hold at all times, for any k ≥ 1, in an optimal
equilibrium. So one might imagine that it would
suffice for the central bank to commit to ensure that
(3.10) holds at all times, where k might be eight
quarters in the future. But if k>1 this condition does
not suffice to determine a unique non-explosive
rational-expectations equilibrium, in the context
of the model set out above. For any commitment of
the form
(3.11)

λ


Et π t +1 + xt +1  = π * +ut ,
κ



where ut is an exogenous random variable satisfying
Et ut + k −1 = 0,
30

J U LY / A U G U S T 2 0 0 4

suffices to determine an equilibrium, for the same
reason that (3.8) does, though the equilibrium will
not be the optimal one (the one determined by (3.8)
except when ut=0 at all times). Yet ensuring that
(3.11) holds at all times is consistent with commitment (3.10); thus all of the different equilibria corresponding to different choices of the process {ut}
are equally consistent with a commitment of the
form (3.10). It follows that a commitment to ensure
(3.10) fails to determine a unique equilibrium, and
indeed it fails to uniquely determine the required
interest-rate policy on the part of the central bank.
A well-known argument for the desirability of
a target criterion referring only to inflation two years
in the future is provided by Svensson (1997). In the
simple model used in that paper for illustrative purposes, the optimal target criterion (in the sense of
Giannoni and Woodford, forthcoming) is shown to be
of this form. But this results because in that model,
an interest-rate decision by the central bank has no
effect on inflation until two years later. It is also true
in the case of the model sketched above, in which
inflation can only be affected by monetary policy
decisions in the previous quarter, that the optimal
target criterion involves a forecast of inflation one
quarter in the future; if the assumed delay were
longer, the optimal target criterion would look farther
into the future.
However, empirical models of the monetary
transmission mechanism do not commonly imply
delays of greater than a quarter before monetary
policy is able to affect inflation, even if (because of
various sorts of inertia in the transmission mechanism) the models imply that the effects of disturbances on the inflation rate are greatest only after
several quarters. For example, the aggregate-supply
relation (3.1) assumed above has the property that
a demand disturbance (due to monetary policy or
some other source) that raises output above its natural rate for several quarters will steadily increase
inflation for several quarters, with the full effect
on inflation being observed only after output has
returned to its natural level. Nonetheless, optimal
policy is described by a target criterion (3.8) that
involves only a one-quarter-ahead inflation forecast.
A similar result is obtained in the more complex
model of Giannoni and Woodford (forthcoming),
discussed in the appendix: The optimal target criteria
involve forecasts at many future horizons, but the
weight is greatest on the forecasts for the nearest
horizon at which the variables in question can still
be affected by the current policy decision.

FEDERAL R ESERVE BANK OF ST. LOUIS

This should not be surprising; if the target criterion is to completely determine a policy decision
at each date, it must specify what defines an acceptable outcome at the nearest date that can still be
influenced by policy, and not merely what must happen later, at dates that can be influenced by later
policy decisions. The preference for “medium term”
target criteria at inflation-targeting central banks
represents a preference for incomplete specifications
of the banks’ policy commitments. This probably
reflects a greater degree of certainty about the desirability of the particular aspect of policy about which
the commitment is being made, and this is understandable. One can indeed state with greater confidence that it is desirable for medium-term inflation
expectations to be highly stable (and to suggest a
plausible value for the target π*) than one can argue
for the desirability of a particular criterion such as
(3.8) that should be satisfied by the transition dynamics for inflation following a temporary disturbance.
Nonetheless, it is possible to make an argument
for a particular near-term target criterion such as
(3.8) that is surprisingly robust. For example, it might
be thought better to leave the transition dynamics
following disturbances unspecified on the grounds
that the optimal transition dynamics will look very
different in the case of different types of disturbances.
Yet Giannoni and Woodford (2002) show that it is
possible quite generally to find a target criterion
that applies regardless of the character of (additive)
disturbances, yet which is sufficiently specific to
uniquely determine the transition dynamics in
response to any type of disturbance.
It is sometimes proposed, in discussions of
inflation-forecast targeting, that a suitable form of
central bank commitment that is specific enough
about the desired transition dynamics to determine
an appropriate policy action involves specification
of a medium-term inflation target, together with a
specification of the rate at which policy should seek
to restore inflation to the target level when it deviates
from it. A commitment of this form can be expressed
in terms of a near-term target criterion of the form
(3.12)

Et π t +1 = π * + µ (π t − π *) ,

where 0<µ<1 indicates the rate at which departures
from the target should be eliminated; thus such a
proposal amounts to a near-term target criterion,
and not simply a medium-term inflation target.
However, it is not generally possible to express a
robustly optimal target criterion (in the sense of
Giannoni and Woodford, 2002) in a form like this—

Woodford

one that makes no reference to the projected path
of any variable other than inflation. A robustly
optimal criterion such as (3.8) implies a particular
rate of convergence of Et π t+k to π* as k is made
large, but this will differ depending on the recent
history of disturbances; it is only the criterion (3.8),
which involves the output-gap projection as well,
that represents a robust criterion for optimality.

3.3 Constant-Interest-Rate Projections
Are an Inappropriate Basis for Policy
One way that inflation-targeting central banks
resolve the problem that their medium-term inflation target alone does not suffice to determine a
particular current interest-rate decision—at least
according to their official rhetoric—is by asking
what constant interest-rate setting over the forecast
horizon would result in a projection consistent with
the medium-term target criterion and then choosing
a current interest-rate operating target at that level.
For example, the Bank of England’s Inflation Reports
justify current policy by showing that a projection
based on the assumption that the interest rate will
remain at the current level for the next two years
indicates projected RPIX inflation equal to 2.5 percent eight quarters from now.25 This does not, however, mean that such banks constrain themselves
to actually maintain a constant interest rate for two
years at a time; instead, a new interest-rate setting
is to be chosen each time the projection exercise is
repeated.26
This “solution” to the problem of the incompleteness of the policy commitment represented
by the medium-term target has the advantage of
being simple to explain to the public—as long as
the public is not sophisticated enough to ask what
it really means—but has a number of unappealing
implications.27 First of all, many optimizing models of the monetary transmission mechanism have
25

Former MPC member Charles Goodhart (2001) describes himself as
having tried to set interest rates in this way, and says “This was, I
thought, what the exercise was supposed to be” (p. 177). Heikensten
(1999) describes the similar procedure used by the Bank of Sweden.

26

Indeed, Goodhart (2001) lists as an advantage of the constant-interestrate projection-based procedure that “no one infers any commitment
from the MPC to abide by that assumption in the future, nor is the
credibility of the MPC damaged when, having made this assumption
in a forecast one month, it decides to change interest rates even in
the next month” (pp. 174-75).

27

Goodhart (2001) reviews what he calls “the prima facie case against”
this approach before offering his defense of it. Other critical discussions
include Leitemo (2003), Svensson (2003b), and Honkapohja and Mitra
(2003).

J U LY / A U G U S T 2 0 0 4

31

REVIEW

Woodford

the property first demonstrated by Sargent and
Wallace (1975) for a rational-expectations IS-LM
framework, namely, that the equilibrium path of
the price level (and hence of the inflation rate) is
indeterminate under the assumption of a fixed
nominal interest rate (or indeed, any exogenously
specified interest-rate process).28 If such a model
were to be used for the central bank’s projection
exercise, the staff would be unable to compute predicted paths for inflation or other variables under
the hypothesis of any constant level of nominal
interest rate, and so unable to assert that one particular level would imply satisfaction of the target
criterion.29
Alternatively, many backward-looking models
(including optimizing models in which expectations
are assumed to be based on extrapolation from past
time series) have the property discussed by Friedman
(1969), namely, that maintaining a constant nominal
interest rate indefinitely will lead to explosive inflation dynamics, through a Wicksellian “cumulative
process.”30 Goodhart (2001) suggests that the Bank
of England’s model has this latter property and that,
as a result, “the rate of change of most variables
visible at the two-year horizon in the Bank’s forecast
generally (though not invariably) tends to persist, and
on occasion to accelerate, in the third and subsequent years” (p. 171). In this case, it is possible to ask
which constant interest rate would imply satisfaction
of the target criterion at a certain finite horizon, but
only at the expense of making it clear that hitting
the target at (say) the eight-quarter horizon does not
also imply expecting to hit it in subsequent quarters.
Hence it cannot be the case that one expects to be
content to maintain the constant-interest-rate policy
indefinitely, even in the absence of any developments
that cannot already be foreseen.31
28

See Woodford (2003, Chap. 4) for further discussion.

29

Leitemo (2003) discusses possible interpretations of the constantinterest-rate projection exercise that would allow it to yield a policy
recommendation even in the case of a forward-looking model of the
transmission mechanism; but these do not eliminate the other unappealing features of such a procedure.

30

See Bullard and Mitra (2002) and Preston (2002) for analyses of forwardlooking models with least-squares learning by the private sector.

31

If one’s model currently implies that inflation will depart significantly
from the target rate at the three-year horizon if interest rates are maintained at their current level for that long, then it also implies that one
should expect that a year from now—barring unforeseen developments—if interest rates have been maintained at their current level,
it will then be forecasted that inflation will depart from the target at
the two-year horizon if interest rates are not changed. Hence one
cannot expect that interest rates should remain at their current level
for an entire year, even in the absence of any “news.”

32

J U LY / A U G U S T 2 0 0 4

In fact, there is no reason to suppose that the
constant interest-rate path represents the bank’s
best current estimate of the future path of interest
rates. This is at least implicitly conceded by the
Bank of England in its published discussions of the
accuracy of its projections.32 In these discussions,
the Bank gives exclusive attention to the projections
that it also publishes in the Inflation Report, in which
an interest-rate path is assumed that corresponds
to current market expectations, rather than to the
projections conditional on the constant interest-rate
path, even though the latter ones are given primary
emphasis in the justification of policy. It is evident
that the Bank does not regard the constant interest
rate assumption as the best available forecast of its
behavior. For if it did, it would want to test the accuracy of the projections made under that assumption,
rather than under whatever contrary assumptions
might be made by traders in financial markets.
Thus the auxiliary assumption that is used to
allow the forecast-targeting procedure to determine
an interest-rate recommendation has the consequence that the targeting procedure is based on
forecasts that are not actually believed, even in the
Bank itself. Such a procedure has the paradoxical
implication that the central bank may choose a
policy under which it does not truly expect the target
criterion to be satisfied, though it may believe that
it would be under the counterfactual hypothesis of
the constant interest rate.
Such a state of affairs can hardly be defended as
conducive to transparency in the conduct of monetary policy. If policy is genuinely based on constantinterest-rate conditional projections, then one’s policy
decisions are not aimed at ensuring satisfaction of
the target criterion that is announced to the public;
and the projections published by the central bank
are not accurate forecasts that should better help
the private sector to correctly anticipate the economy’s evolution. On the other hand, if the central
bank genuinely does expect the target criterion to
be satisfied, then policy is not actually determined
in the way that the official rhetoric implies that it
is; and if the forecasts are unbiased, then they are
not the kind of forecasts that they are officially
described as being.
The kind of forecast-targeting procedure recommended by Svensson and Woodford (forthcoming)
as a way of implementing optimal monetary policy
32

See the Bank of England’s Inflation Reports of August 2001 and
August 2002.

FEDERAL R ESERVE BANK OF ST. LOUIS

is of a different sort. In this procedure, one projects
the economy’s future evolution under alternative
contemplated policy decisions, assuming that in
future decision cycles the central bank will again
act to ensure satisfaction of the target criterion. This
amounts to asking what action is needed to project
that the criterion should be satisfied in the current
period, taking as given that it is expected to be satisfied in later periods (as a result of the policy actions
to be taken in those periods). Such a calculation
yields a determinate outcome as long as there is a
determinate rational-expectations equilibrium
implied by the target criterion; this is always the
case if the target criterion is selected according to
the method of Giannoni and Woodford (2002).
Thus policy should be based on a projection
exercise that includes a model of the central bank’s
own future behavior—one that is furthermore consistent with the procedure that it actually follows
in making its policy decisions. This is the kind of
projection exercise used as the basis for policy decisions at some central banks, notably the Reserve
Bank of New Zealand, which also publishes some
information about the non-constant interest-rate
path implicit in its projections, along with its projections for inflation and other variables.
Goodhart (2001) objects that such a procedure
is impractical, on the grounds that it would be much
more difficult for a monetary policy committee to
reach agreement on an entire future path for interest
rates than to decide only about the current interest
rate each time they meet. But the procedure described
by Svensson and Woodford (forthcoming) does not
involve a multidimensional decision problem in
each decision cycle. As with the constant-interestrate projection method, one makes a decision for
the current period only, on the basis of projections
of the future that (necessarily) incorporate a hypothesis about future policy; the hypothesis about future
policy is simply a more realistic one than the notion
that interest rates will not change, regardless of how
inflation and output evolve. And there is no greater
need for agreement among the members of the
policy committee about that particular aspect of the
model specification than about the other assumptions involved in making projections for the future.
Goodhart (2001) also argues that revealing a
projected non-constant path for interest rates is
problematic, because “any indication that the MPC
is formally indicating a future specific change in
rates...would be taken to indicate some degree of
commitment” (p. 175). This is clearly a delicate issue

Woodford

regarding the proper explanation and the public’s
subsequent interpretation of the central bank’s
projections. Yet the experience in New Zealand suggests that it is possible to reveal interest-rate projections to the public without being understood to have
made an advance commitment about the path of
the official cash rate. Moreover, a “fan chart” for the
path of interest rates, like those that the Bank of
England currently publishes for its inflation and
output projections, ought to make it clear that the
bank is not committing itself to a definite path;
rather, the expected evolution will depend on a
variety of contingencies that can at best be assigned
probabilities.
If necessary for the reasons to which Goodhart
(2001) refers, it would be preferable to base policy on
projections conditioned on predicted future policy,
and to publish inflation and output projections of
that sort, without any mention of the interest-rate
path implicit in these projections, than to base policy
on projections conditional upon a model of policy
that one knows to be false. But there are likely to
be advantages to publication of the interest-rate
projections. One of the crucial ways in which central
banks affect the economy is through the effects that
their announcements have on expectations regarding
the future path of short-term interest rates, expectations that then determine longer-term bond yields,
asset prices, and exchange rates, which in turn affect
spending, employment, wage-setting, and pricesetting decisions. The current level of overnight
interest rates is in itself of little importance for most
economic decisions; the real significance of central
bank decisions about the overnight rate is what they
are taken to signal about the likely path of interest
rates months and years into the future. Given the
importance to a central bank of steering expectations
of future interest rates in a desirable way, it would
seem that revealing to the public the expected future
path of rates implied by the bank’s policy commitments should help it to better achieve its goals.

3.4 Advantages of a HistoryDependent Target Criterion
A notable feature of the kind of projection
exercises upon which policy is currently based at
banks like the Bank of England is that they are purely
forward-looking. By this I mean that the decision
made at any time is a function solely of the policy
committee’s judgment about the possible paths
from now on for inflation and other variables (if
J U LY / A U G U S T 2 0 0 4

33

Woodford

any) relevant to its target criterion; past conditions
are irrelevant except insofar as these have an effect
on what it is possible to achieve from now on. Of
course any projection-based decision procedure
will be forward-looking; but under a procedure like
the Bank of England’s (at least as Goodhart, 2001,
describes it), the past is irrelevant because the target
criterion is a time-invariant function of the projected
future path of the target variable (RPIX inflation).
One might think that forward-looking behavior
of this kind is a necessary feature of optimal policy—
that “bygones should be bygones” for a rigorous
optimizer. But as explained above, this is not correct
in the case of the optimal control of a forwardlooking system. If it were, there would be no flaw
in the reasoning of a purely discretionary policymaker. When the private sector is forward-looking,
expectations regarding future policy matter for what
can be achieved at any point in time, and outcomes
can generally be improved through a judicious commitment regarding future policy. This requires, however, that policy be expected to be conducted at the
later date in a way that is history-dependent—that
is, in a way that depends on the earlier conditions
(at the time at which it was desirable to alter expectations) as well as upon conditions at the time that
the action is taken.
This history-dependence can be incorporated
into a forecast-targeting procedure through the use
of a history-dependent target criterion to evaluate
whether the economy’s projected evolution from
now on should be considered to be consistent with
the bank’s general policy commitments. This means
that the acceptable projections for the target variables
looking forward should depend on recent past conditions.33 This is a further reason why, under an
optimal regime, the short-term target for inflation
will be time-varying, even though there is likely to
be a constant long-run inflation target, around which
the short-term target fluctuates. The way in which
33

The optimal target criterion (3.8) in the simple example above might
seem not to confirm this principle, as it involves only forecasts of
π t+1 and xt+1. But in that model, inflation is not technically a “target
variable,” because it is the rate of inflation acceleration, rather than
the absolute rate of inflation, that enters the loss function (3.2). A purely
forward-looking target criterion would then be one that involves only
the projected future paths of the output gap and of inflation acceleration.
The target criterion (3.8) is not of this form, as it implies a commitment to eventually reverse past increases in the inflation rate. We could
alternatively adopt (3.6) as a target criterion, and this would also be
optimal. This criterion involves only the projected acceleration of
inflation in period t+1, not the absolute rate of inflation. However,
the criterion is history-dependent because of its dependence on the
value of E t –1xt .

34

J U LY / A U G U S T 2 0 0 4

REVIEW
an optimal short-term target criterion is likely to be
history-dependent is illustrated in the discussion
in the appendix (of the optimal target criteria in
the case of the estimated model of Giannoni and
Woodford, forthcoming).
A particularly clear example of the advantages
of a history-dependent target criterion is the situation
currently faced by the Bank of Japan, in which the
zero lower bound on nominal interest rates has been
reached and yet deflation continues, so that further
monetary stimulus is desirable. As I have already
mentioned, the main lever by which monetary policy
can still affect the economy under such circumstances is by changing expectations regarding the
future conduct of policy: Committing to a more
expansionary policy later than would otherwise have
been pursued. But this requires that policy not be
expected to be conducted later in accordance with
a purely forward-looking target criterion. For example, Eggertsson and Woodford (2003) show that the
expectation that the central bank will remain committed to the forward-looking pursuit of a low (timeinvariant) inflation target—and hence will adjust
interest rates so as to be consistent with the target
as soon as this can be done without violation of the
zero lower bound — can lead to a disastrous outcome
when real disturbances result in a temporarily negative natural rate of interest. This analysis suggests
that the problem of the Bank of Japan at present is
not that it is not understood to be committed to a
non-negative inflation target, but that it is expected
to pursue its tacit inflation target in a purely forwardlooking manner, with the implication that the
(unwanted) price declines that occur while the zero
bound constrains policy will never be undone.
A commitment to a time-invariant inflation target
would be more likely to avoid the problem caused
by the zero bound, of course, if the target were set
several percentage points above zero, as advocated
by Summers (1991). But this would result in substantial losses of another sort, those created by chronic
inflation. The optimal policy rule, as Eggertsson and
Woodford show, would instead involve commitment
to a history-dependent target criterion, resulting in
temporary inflation after a period in which the zero
bound constrained policy; in addition, a greater
amount of inflation would occur the longer the zero
bound continued to bind and the greater the cumulative deflation that occurred during that time. A low
inflation rate would again be targeted once a sufficient period of time passed in which the zero bound

FEDERAL R ESERVE BANK OF ST. LOUIS

did not prevent the central bank from hitting its
target. Credible commitment to a history-dependent
policy of this kind would create the desired kind of
expectations while the economy is in the “liquidity
trap”—so that the deflation and output contraction
at that time should remain quite modest—without
requiring chronic inflation during normal times and
creating an “inflation scare” during the period in
which the economy is reflated as it exits from the
trap.

4. CONCLUSIONS
Inflation-forecast targeting, as currently practiced
at central banks such as the Bank of England, represents an important innovation in decision procedures with regard to monetary policy, one that has
moved the actual practice of leading central banks
closer to the ideal that would be recommended on
the basis of economic theory. The organization of
the decision process around the achievement of an
explicit, quantitative target that is also communicated
to the public, and a commitment to the explanation
of policy decisions to the public in terms that allow
verification of the central bank’s commitment to its
putative target are important improvements upon
prior procedures. They can both help to safeguard
a central bank against the trap of discretionary
policymaking, and help the private sector to more
accurately anticipate future policy, increasing the
effectiveness of policy. The introduction of targeting
rules as a way of specifying policy commitments is
also an important conceptual advance, allowing
commitments to be stated in a way that incorporates
a kind of flexibility that is of considerable practical
value, while being specific about the aspects of
policy that are most critical for anchoring privatesector expectations.
At the same time, current practice falls short of
the theoretical ideal sketched in this paper in some
notable respects. Perhaps the most important of
these is the exclusive emphasis on “medium term”
targets that leave unspecified the basis on which a
particular nearer-term path toward that target is to
be preferred. At best, this represents a significant
degree of vagueness about the criterion that is
actually used to make policy decisions. It may also
indicate that the choice among alternative near-term
paths for the economy is still made on a discretionary basis that will ensure suboptimal policy even
when decisions are made by an omniscient monetary policy committee with a perfect understanding

Woodford

of social welfare. The question whether it would be
practical for central banks to commit themselves
to more explicit nearer-term target criteria, of the
form indicated by the theory of optimal monetary
policy rules, should be an important issue for further
study by central bankers and monetary economists
alike.

REFERENCES
Altig, David; Christiano, Lawrence J.; Eichenbaum, Martin
S. and Linde, Jesper. “Technology Shocks and Aggregate
Fluctuations.” Unpublished manuscript, Federal Reserve
Bank of Cleveland, 2002.
Amato, Jeffery D. and Laubach, Thomas. “Estimation and
Control of an Optimization-Based Model with Sticky
Prices and Wages.” Journal of Economic Dynamics and
Control, May 2003, 27(7), pp. 1181-215.
Aoki, Kosuke. “Optimal Monetary Policy Responses to
Relative Price Changes.” Journal of Monetary Economics,
August 2001, 48(1), pp. 55-80.
Auerbach, Alan J. and Obstfeld, Maurice. “The Case for
Open-Market Purchases in a Liquidity Trap.” NBER
Working Paper No. 9814, National Bureau of Economic
Research, July 2003.
Ball, Laurence and Sheridan, Niamh. “Does Inflation
Targeting Matter?” in Ben S. Bernanke and Michael
Woodford, eds., The Inflation Targeting Debate. Chicago:
University of Chicago Press (forthcoming).
Ball, Laurence; Mankiw, N. Gregory and Reis, Ricardo.
“Monetary Policy for Inattentive Economies.” NBER
Working Paper No. 9491, National Bureau of Economic
Research, February 2003.
Barro, Robert J. and Gordon, David B. “A Positive Theory of
Monetary Policy in a Natural Rate Model.” Journal of
Political Economy, August 1983, 91(4), pp. 589-610.
Benigno, Pierpaolo. “Optimal Monetary Policy in a
Currency Area.” Journal of International Economics
(forthcoming).
Bernanke, Ben S.; Laubach, Thomas; Mishkin, Frederic S.
and Posen, Adam S. Inflation Targeting: Lessons from
International Experience. Princeton: Princeton University
Press, 1999.
Blinder, Alan; Goodhart, Charles; Hildebrand, Philipp; Lipton,

J U LY / A U G U S T 2 0 0 4

35

Woodford

David and Wyplosz, Charles. How Do Central Banks Talk?
Geneva Report on the World Economy No. 3. London: Centre
for Economic Policy Research/International Center for
Monetary and Banking Studies, 2001.

REVIEW
Goodfriend, Marvin. “Inflation Targeting in the United
States?” in Ben S. Bernanke and Michael Woodford, eds.,
The Inflation Targeting Debate. Chicago: University of
Chicago Press, (forthcoming).

Bullard, James B. and Mitra, Kaushik. “Learning about
Monetary Policy Rules.” Journal of Monetary Economics,
September 2002, 49(6), pp. 1105-129.

Goodfriend, Marvin and King, Robert G. “The New
Neoclassical Synthesis and the Role of Monetary Policy.”
NBER Macroeconomics Annual, 1997, 12(1), pp. 231-83.

Calvo, Guillermo A. “Staggered Prices in a Utility-Maximizing
Framework.” Journal of Monetary Economics, September
1983, 12(3), pp. 383-98.

Heikensten, Lars. “The Riksbank’s Inflation Target—
Classification and Evaluation.” Bank of Sweden Quarterly
Review, 1999, 1, pp. 5-17.

Christiano, Lawrence J.; Eichenbaum, Martin and Evans,
Charles L. “Nominal Rigidities and the Dynamic Effects
of a Shock to Monetary Policy.” NBER Working Paper
No. 8403, National Bureau of Economic Research, July
2001.

Honkapohja, Seppo and Mitra, Kaushik. “Problems in
Inflation Targeting Based on Constant Interest Rate
Projections.” Unpublished manuscript, University of
Helsinki, August 2003.

Eggertsson, Gauti B. and Woodford, Michael. “The Zero
Bound on Interest Rates and Optimal Monetary Policy.”
Brookings Papers on Economic Activity, 2003, (1), pp.
139-211.
Erceg, Christopher J.; Henderson, Dale W. and Levin,
Andrew T. “Optimal Monetary Policy with Staggered
Wage and Price Contracts.” Journal of Monetary Economics,
October 2000, 46(2), pp. 281-313.
Friedman, Milton. “The Optimum Quantity of Money,” in
Milton Friedman, ed., The Optimum Quantity of Money
and Other Essays. Chicago: Aldine, 1969.
Fuhrer, Jeffrey C. and Moore, Geoffrey R. “Inflation
Persistence.” Quarterly Journal of Economics, February
1995, 110(1), pp. 127-59.
Giannoni, Marc P. and Woodford, Michael. “Optimal
Interest-Rate Rules: I. General Theory.” NBER Working
Paper No. 9419, National Bureau of Economic Research,
December 2002.
Giannoni, Marc P. and Woodford, Michael. “Optimal
Inflation Targeting Rules,” in Ben S. Bernanke and
Michael Woodford, eds., The Inflation Targeting Debate.
Chicago: University of Chicago Press (forthcoming).
Goodhart, Charles A.E. “Monetary Transmission Lags and
the Formulation of the Policy Decision on Interest Rates.”
Federal Reserve Bank of St. Louis Review, July/August
2001, 83(4), pp. 165-81.
36

J U LY / A U G U S T 2 0 0 4

Khan, Aubhik; King, Robert G. and Wolman, Alexander L.
“Optimal Monetary Policy.” NBER Working Paper No. 9402,
National Bureau of Economic Research, December 2002.
King, Mervyn. “What Has Inflation Targeting Achieved?”
in Ben S. Bernanke and Michael Woodford, eds., The
Inflation Targeting Debate. Chicago: University of Chicago
Press (forthcoming).
King, Robert G. and Wolman, Alexander L. “What Should
the Monetary Authority Do When Prices Are Sticky?” in
John B. Taylor, ed., Monetary Policy Rules. Chicago:
University of Chicago Press, 1999.
Krugman, Paul. “It’s Baaack: Japan’s Slump and the Return
of the Liquidity Trap.” Brookings Papers on Economic
Activity, 1998, (2), pp. 137-87.
Kydland, Finn E. and Prescott, Edward C. “Rules Rather
than Discretion: The Inconsistency of Optimal Plans.”
Journal of Political Economy, 1977, 85(3), pp. 473-91.
Lange, Joe; Sack, Brian and Whitesell, William. “Anticipations
of Monetary Policy in Financial Markets.” Finance and
Economic Discussion Paper No. 2001-24, Board of
Governors of the Federal Reserve System, May 2001.
Leiderman, Leonardo and Svensson, Lars E.O., eds. Inflation
Targets. London: Centre for Economic Policy Research,
1995.
Leitemo, Kai. “Targeting Inflation by Constant-Interest-Rate
Forecasts.” Journal of Money, Credit and Banking, August
2003, 35(4), pp. 609-26.

FEDERAL R ESERVE BANK OF ST. LOUIS

Orphanides, Athanasios and Williams, John C. “Imperfect
Knowledge, Inflation Expectations, and Monetary Policy,”
in Ben S. Bernanke and Michael Woodford, eds., The
Inflation Targeting Debate. Chicago: University of Chicago
Press (forthcoming).
Preston, Bruce. “Learning about Monetary Policy Rules
when Long-Horizon Forecasts Matter.” Unpublished
manuscript, Princeton University, August 2002.
Rotemberg, Julio J. and Woodford, Michael. “An OptimizationBased Econometric Framework for the Evaluation of
Monetary Policy,” in Ben S. Bernanke and Julio J.
Rotemberg, eds., NBER Macroeconomics Annual 1997.
Volume 12, No. 1. Cambridge, MA: MIT Press, 1997, pp.
297-346.
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.
Smets, Frank and Wouters, Raf. “Monetary Policy in an
Estimated Stochastic Dynamic General Equilibrium
Model of the Euro Area.” Unpublished manuscript,
European Central Bank, May 2002a.

Woodford

Svensson, Lars E.O. “Inflation Targeting as a Monetary
Policy Rule.” Journal of Monetary Economics, June 1999,
43(3), pp. 607-54.
Svensson, Lars E.O. “What Is Wrong with Taylor Rules?
Using Judgment in Monetary Policy Through Targeting
Rules.” Journal of Economic Literature, June 2003a, 41(2),
pp. 426-77.
Svensson, Lars E.O. “The Inflation Forecast and the Loss
Function,” in P. Mizen, ed., Central Banking, Monetary
Theory and Practice: Essays in Honour of Charles Goodhart,
Vol. 1. Edward Elgar, 2003b, pp. 135-52.
Svensson, Lars E.O. and Woodford, Michael. “Implementing
Optimal Policy Through Inflation-Forecast Targeting,” in
Ben S. Bernanke and Michael Woodford, eds., The
Inflation Targeting Debate. Chicago: University of Chicago
Press (forthcoming).
Vickers, John. “Inflation Targeting in Practice: The U.K.
Experience.” Bank of England Quarterly Bulletin,
November 1998, 38(4), pp. 368-75.
Woodford, Michael. “The Optimum Quantity of Money,” in
B.M. Friedman and F.H. Hahn, eds., Handbook of Monetary
Economics. Volume II. Amsterdam: North-Holland, 1990.

Smets, Frank and Wouters, Raf. “Sources of Business Cycle
Fluctuations in the U.S.: A Bayesian DSGE Approach.”
Seminar presentation, Princeton University, November 1,
2002b.

Woodford, Michael. “Optimal Monetary Policy Inertia.”
NBER Working Paper No. 7261, National Bureau of
Economic Research, July 1999a.

Summers, Lawrence. “How Should Long-Term Monetary
Policy Be Determined?” Journal of Money, Credit, and
Banking, August 1991, 23(3), pp. 625-31.

Woodford, Michael. “Commentary: How Should Monetary
Policy Be Conducted in an Era of Price Stability?” in New
Challenges for Monetary Policy. Kansas City: Federal
Reserve Bank of Kansas City, 1999b.

Svensson, Lars E.O. “Inflation Forecast Targeting:
Implementing and Monitoring Inflation Targets.”
European Economic Review, June 1997, 41(6), pp. 1111-46.

Woodford, Michael. Interest and Prices: Foundations of a
Theory of Monetary Policy. Princeton: Princeton University
Press, 2003.

J U LY / A U G U S T 2 0 0 4

37

REVIEW

Woodford

Appendix

AN OPTIMAL TARGETING RULE
FOR THE MODEL OF GIANNONI AND
WOODFORD (FORTHCOMING)
Here I summarize the quantitative form of the
optimal targeting rule derived by Giannoni and
Woodford (forthcoming) in the context of a small,
empirical optimizing model of the U.S. monetary
transmission mechanism. The desirability of this
precise rule depends, of course, on the details of
the quantitative model, many of which are highly
debatable. Nonetheless, it may be useful to consider
this example of an optimal policy rule for an estimated model, as an illustration of some of the general
points made in the text about the likely character of
an optimal policy rule.
The model of Giannoni and Woodford incorporates both wage and price stickiness, with random
intervals between the times at which both individual
wages and prices are reconsidered, as in the theoretical analysis of Erceg, Henderson, and Levin (2000).
In addition, both wages and prices are allowed to
be indexed to the previous quarter’s index of
prices between the occasions on which they are reoptimized, as proposed by Christiano, Eichenbaum,
and Evans (2001). The degrees of indexation of both
wages and prices are treated as free parameters to
be estimated, as in Smets and Wouters (2002a,
2002b), but our parameter estimates indicate the
best fit under the assumption of full indexation of
both wages and prices, as assumed by Christiano,
Eichenbaum, and Evans. Both wages and prices are
also determined a quarter in advance. On the
demand side of the model, the preferences of the
representative household are assumed to allow for
habit persistence, and the best fit is obtained when
the habit-persistence coefficient takes the largest
allowable value, so that utility depends on the change
in real expenditure rather than its level. In addition,
real private expenditure is determined two quarters
in advance. The several free parameters of the model
are estimated by minimizing the distance between
the predicted impulse responses of four variables
(output, inflation, the real wage, and the short-term
nominal interest rate) to a monetary policy shock
and those implied by an unrestricted VAR model of
the same four time series. The parameter estimates
are consistent with the sign restrictions implied by
theory, and several restricted versions of the model—
a restricted model with no indexation to the lagged
38

J U LY / A U G U S T 2 0 0 4

price index, a restricted model with flexible wages,
and a restricted model with no habit persistence—
can each be statistically rejected.
Here I summarize the implications for optimal
policy of treating the best-fitting parameter values
as representing the literal truth. First of all, the estimated model implies that maximization of the
expected utility of the representative household
corresponds to minimization of a quadratic loss
function of the form
(A.1)

(

)

λ π − γ π 2
p
t
p t −1
E0 ∑ β t 
w
t =0
+ λw π t − γ wπ t −1
`

(


,
2
2
+ λ x ( xt − δxt −1 − xˆ *) 


)

where πt is an index of goods price inflation between
quarter t –1 and quarter t, πtw is an index of wage
inflation, and xt is the output gap (log real output
relative to a “natural rate of output” that varies in
response to several types of real disturbances). The
discount factor is calibrated to equal 0.99 (to imply
a realistic long-run average real rate of return), while
the model’s estimated parameters imply the values
λp=0.9960, λw=0.0040, λx=0.0026, and δ=0.035
for the coefficients of the loss function.
The fact that prices are indexed to a lagged price
index implies that it is inflation acceleration, rather
than the rate of inflation as such, that creates distortions, as in the simpler model discussed in the text.
The fact that wages are also sticky implies that wage
inflation also creates distortions, even when the rate
of goods price inflation is stable; because wages
are indexed to the lagged price index, it is actually
wage inflation relative to lagged price inflation that
measures this distortion. Finally, because of habit
persistence, the distortions associated with fluctuations in the output gap are not proportional simply
to a sum of squared deviations of the output gap each
period from its optimal level, but rather to a sum of
squared deviations of the output gap from an increasing function of the previous quarter’s output gap.
However, the weight δ on the lagged output gap turns
out to be quite small, despite the existence of substantial habit persistence.
We also find that the estimated parameter values
imply a very small relative weight on the wageinflation stabilization objective relative to the priceinflation stabilization objective. This is not because
wages are found to be flexible, but because other

FEDERAL R ESERVE BANK OF ST. LOUIS

estimated parameters imply larger distortions
resulting from misalignment of prices than from
misalignment of wages. The relative weight on the
output-gap stabilization objective implied by the
parameter estimates is also quite small; this follows
directly from the estimation of parameters that imply
only weak responses of wage and price inflation to
variations in the output gap, as discussed in the text.
An optimal policy for the estimated model—
and one with the desirable property that it is optimal
regardless of the assumed statistical properties of the
disturbances, and not solely in the case of disturbance processes of the kind implied by the estimated
model for the historical sample period—can be
implemented by a targeting procedure of the following kind.34 First, in each quarter t, the central bank
intervenes in the money markets (through open
market operations, repurchases, standing facilities
in the interbank market for central bank balances,
etc.) so as to implement the interest rate target it,t–1
announced in quarter t–1. As in the simpler model
discussed in the text, the fact that wages, prices, and
spending are all predetermined for a quarter implies
that nothing can be gained from allowing variations
in interest rates that are not forecastable in the previous quarter.
Second, in the quarter-t decision cycle, the bank
must choose an operating target it+1,t to announce
for the following quarter. This is chosen in order to
imply a projected evolution of (wage and price)
inflation from quarter t+1 onward that satisfies a
target criterion of the form

Woodford

forecasts at different horizons, k. We observe that
the target criterion can be thought of as a wageadjusted inflation target.
Third, it is also necessary, as part of the quarter-t
decision cycle, for the central bank to choose the
target π–t+1 for the following quarter. This is chosen
so as to ensure that future policy will be conducted
in a way that allows the bank to project (conditional
on its current information) that another target criterion, of the form
Ft* (π ) + φ*w Ft* (w) + φ*x Ft* ( x ) = π t* ,

(A.4)

should be satisfied, where the expressions Ft*(z) are
again weighted averages of forecasts at different
horizons (but with relative weights αkz* that may be
different in this case) and π t* is another time-varying
target value, once again a predetermined variable.
In this case the criterion specifies a target for a wageand output-adjusted inflation projection.
In this last procedure, optimality requires that
the target value be given by an expression of the
form
(A.5)

(

)

π t* = 1 − θ π* π * +θ π* Ft1−1(π ) + θ *x Ft1−1(w) + θ *x Ft1−1( x ) ,

sum to 1. Thus the coefficient
where the weights
φw is actually the sum of the weights on real-wage

where the expressions F1t (z) are still other weighted
averages of forecasts at different horizons, with relative weights α kz1 that again sum to 1, and π* is an
arbitrary constant.35 Note that the optimal target
value depends on the previous quarter’s forecasts
of the economy’s subsequent evolution; this is an
example of the history dependence of optimal target
criteria, discussed generally in the text.
The estimated parameter values imply the following numerical coefficients in the optimal target
criteria. In the case of the short-term criterion (A.2),
the coefficient φw is equal to 0.565.36 Thus if unexpected developments in quarter t are projected to
imply a higher future level of real wages than had
previously been anticipated, policy must ensure that
projected future price inflation is correspondingly
reduced. This is because of a desire to stabilize
(nominal) wage inflation as well as price inflation,

34

35

Note that in the model considered here, as in the simpler model discussed in the text, there is no welfare significance to any absolute
inflation rate, only to changes in the rate of inflation and to wage
growth relative to prices. There is therefore no particular inflation
rate that could be justified as optimal from a timeless perspective.

36

Here and below, the coefficients are presented for a target criterion
where the inflation rate is measured in annualized percentage points.

(A.2)

[

]

Ft (π ) + φw Ft (w) − wt = π t ,

where π–t is a target value that has been determined
in quarter t–1. Here for each of the variables z=π,w,
the expression Ft (z) refers to a weighted average of
forecasts of the variable z at various future horizons,
conditional on information at date t:
`

(A.3)

Ft ( z ) ; ∑ α kz Et z t + k ,
k =1

α kz

Because the empirical model is quarterly, it is simplest to discuss the
policy process as if a policy decision is also made once per quarter,
even though in reality most central banks reconsider their operating
targets for overnight interest rates somewhat more frequently than
this. The discussion should not be taken to imply that it is optimal
for the policy committee to meet only once per quarter; this would
follow only if (as in the model) all other markets were also open only
once per quarter.

J U LY / A U G U S T 2 0 0 4

39

REVIEW

Woodford

the current target π–t–1). This is a criterion in the spirit
of inflation-forecast targeting as currently practiced
at central banks such as the Bank of England, except
that projected wage growth matters as well as price
inflation, and that the target shifts over time.
In the case of the long-term criterion (A.4),
instead, the numerical coefficients of the target
criterion are given by

Figure A1
α πk

αw
k

0.25

1.2
1

0.2
0.8
0.15

0.6

φ*w = 0.258,

0.4

0.1

0.2
0.05
0
0

–0.2
1

2

3

4

5

6

1

2

3

4

5

6

NOTE: Relative weights on projections at different horizons in
the short-run target criterion (A.2). The horizontal axis indicates
the horizon k in quarters.

and under circumstances of expected real wage
growth, inflation must be curbed in order for nominal wage growth to not be even higher.
The relative weights that this criterion places
on projections at different future horizons are shown
in Figure A1. The two panels plot the coefficients
α kπ and α kw as functions of the horizon k. Note that
in each case the quarter for which the projections
receive greatest weight is one quarter in the future.
This is also the first quarter in which it is possible for
wage or price inflation to be affected by the choice
of it+1,t , according to the estimated model. However,
while the real-wage projection that matters is primarily the projected growth in real wages between
the present quarter and the next one, substantial
weight is also placed on projected inflation farther
in the future; in fact, the mean lead Σkα kπk is between
10 and 11 quarters in the future in the case of the
inflation projection Ft (π ). Thus the short-run target
criterion is a (time-varying) target for the average rate
of inflation that is projected over the next several
years, adjusted to take account of expected wage
growth, mainly over the coming quarter. Roughly
speaking, optimal policy requires the central bank
to choose Et it+1 in quarter t to head off any change
in the projected average inflation rate over the next
several years that is due to any developments not
anticipated in quarter t –1 (and hence reflected in
40

J U LY / A U G U S T 2 0 0 4

φ*x = 0.135.

In this case, output-gap projections matter as
well; a higher projected future output gap will require
a reduction in the projected future rate of inflation,
just as will a higher projected future real wage. The
numerical size of the weight placed on the outputgap projection may appear modest; but as we shall
see below, the degree of variability of output-gap
projections in practice is likely to make this a quite
significant correction to the path of the target
criterion.
The relative weights on forecasts at different
horizons in this criterion are plotted in the panels
in the first row of Figure A2. We observe that in the
case of this criterion, the projections that mainly
matter are those for two quarters in the future; the
criterion is nearly independent of projections regarding the quarter after the current one. Hence it makes
sense to think of this criterion as the one that should
determine the central bank’s intended policy two
or more quarters in the future (and hence its choice
in quarter t of the target π–t+1 to constrain its choice
in the following period of i t+2,t+1); but this criterion
should not be thought of as a primary determinant
of whether the bank’s intended policy in period t+1
is on track. The projections that receive the greatest
weight under this criterion are those for the same
quarter (quarter t+2) that will receive the greatest
weight in the targeting procedure for which π–t+1
provides the target value.
Finally, the coefficients of the rule (A.5) determining the target value for the long-term criterion
are given by

θ π* = 0.580,

θ w* = 0.252,

θ *x = 0.125 .

The weights in the projections (conditional on
information in the previous quarter) at various
horizons are plotted in the second row of Figure A2.
Here, too, it is primarily projections for two quarters
in the future that matter in each case. Roughly speaking, then, the target value for the wage- and output-

FEDERAL R ESERVE BANK OF ST. LOUIS

Woodford

Figure A2
α*k π

α*w
k

α*x
k

0.35

1.2

1.2

0.3

1

1

0.25

0.8

0.8

0.2

0.6

0.6

0.15

0.4

0.4

0.1

0.2

0.2

0.05

0

0

0

2

4

6

–0.2

2

α kπ 1

4

6

–0.2

α kw1
1.2

1.2

0.5

1

1

0.4

0.8

0.8

0.3

0.6

0.6

0.2

0.4

0.4

0.1

0.2

0.2

0

0

0

2

4

6

–0.2

2

4

6

4

6

α kx1

0.6

–0.1

2

4

6

–0.2

2

NOTE: Relative weights on projections at different horizons in the long-run target criterion. Panels in
the first row indicate the projections in (A.4), while the second row indicates the projections from the
previous quarter that define the target value π *t .

adjusted inflation projection two quarters in the
future is high when a similar adjusted inflation
projection (again, for a time two quarters in the
future) was high in the previous quarter.
Thus forecasting exercises, in which the central
bank projects the evolution of both inflation and
real variables many years into the future under alternative hypothetical policies on its own part, play a
central role in a natural approach to the implementation of optimal policy. A forecast of inflation several
years into the future is required in each (quarterly)
decision cycle in order to check whether the intended
interest-rate operating target for the following quarter
is consistent with the criterion (A.2). In addition, the
time-varying medium-term inflation target π–t must
be chosen each period on the basis of yet another
forecasting exercise. While the long-run target crite-

rion (A.4) primarily involves projections for a time
only two quarters in the future, the choice of π–t+1
requires that the central bank solve for a projected
path of the economy in which (A.4) is satisfied not
only in the current period, but in all future periods
as well. Hence this exercise as well requires the construction of projected paths for inflation and real
variables extending many years into the future. The
relevant paths, however, will not be constant-interestrate projections, but rather projections of the economy’s future evolution given how policy is expected
to evolve. Indeed, the projections are used to select
constraints upon the bank’s own actions in future
decision cycles, by choosing both the interest-rate
operating target it+1,t and the adjusted inflation target
π–t+1 in period t.

J U LY / A U G U S T 2 0 0 4

41

REVIEW

42

J U LY / A U G U S T 2 0 0 4

Commentary
Stephanie Schmitt-Grohé

W

oodford concludes his review of what the
theoretical literature on optimal monetary policy has to say about the desirability of price stability with the following statement:
“It is not a bad first approximation to say that the
goal of monetary policy should be price stability.”
It follows from this conclusion that the findings of
the theoretical literature on optimal monetary policy
can be interpreted as supportive of inflation targeting. However, it does not imply that there should
be a single target variable—namely, inflation.
Optimal policy, as Woodford explains, can only
under quite special circumstances be described
solely in terms of the behavior of inflation.
Woodford argues that even when full price stability fails to be optimal, near price stabilization is
optimal in many cases. In particular, Woodford
discusses that near stabilization of an appropriately
defined price index continues to be optimal (i) in
environments where negative real rates in combination with the zero bound on the nominal interest
rate make a path of zero inflation impossible, (ii) in
environments where asymmetric shocks require
relative price changes, (iii) in sticky-wage models, and
(iv) even in cases where the flexible-price equilibrium
is not efficient, due to the presence of (possibly timevarying) market power or distorting taxes.
I would like to expand on this discussion and
add to the list of environments in which near price
stability is optimal. Also, I would like to give some
more specific examples. Most of the theoretical work
Woodford surveys uses dynamic, stochastic general
equilibrium models that contain some specific simplifying assumption that make it possible to accurately characterize optimal policy using only linear
approximations to the model and that allow for an
analytical characterization of optimal policy. Absent
those simplifying assumptions, one would have to
use higher-order approximations to the equilibrium
conditions for welfare calculations. Recent advances

in computational economics have delivered algorithms that make it feasible and simple to compute
higher-order approximations to the equilibrium
conditions of a general class of large stochastic
dynamic general equilibrium models (see, for
instance, Sims, 2000, and Schmitt-Grohé and Uribe,
2004a). Several authors have applied this toolkit to
studying the welfare consequences of monetary
policy in environments without the special assumption needed to make the linear approach work. And
I will report findings on the desirability of price
stability from this more numerically oriented branch
of the literature.
Academic economists in the past have not
always arrived at the conclusion that price stability
should be a central objective of monetary policy. In
particular, many of the theoretical environments
that were used to study optimal policy in the 1980s
assumed that there are no impediments to instantaneous adjustment of factor and product prices. In
such environments, price stability in the sense of a
constant price level over time does not in general
represent the optimal monetary policy prescription.
Rather, under optimal monetary policy, prices move
over time in such a way as to eliminate the opportunity cost of holding money, that is, optimal policy
follows the Friedman rule. The opportunity cost of
holding money is the nominal interest rate, and, thus,
optimal monetary policy calls for a constant and
zero nominal interest rate. With nominal interest
rates constant, prices move in response to changes
in real interest rates, and fall on average at the real
rate of interest.
In addition, Chari, Christiano, and Kehoe (1991)
show that in a world in which the Friedman rule is
optimal, optimal monetary policy is associated with
high inflation volatility. In Chari, Christiano, and
Kehoe, the reason why inflation is highly volatile
under the optimal policy is that the government is
using surprise inflation as a non-distorting fiscal

Stephanie Schmitt-Grohé is a professor of economics at Duke University.
Federal Reserve Bank of St. Louis Review, July/August 2004, 86(4), pp. 43-49.
© 2004, The Federal Reserve Bank of St. Louis.

J U LY / A U G U S T 2 0 0 4

43

REVIEW

Schmitt-Grohé

Table 1
Desirability of Price Stability in an Optimal
Monetary and Fiscal Policy Problem
Variable

Mean

Standard
deviation Autocorrelation

Flexible-price economy

π
R

–3.66
0

6.04

–0.04

0

—

Sticky-price economy

π

–0.16

0.17

0.04

R

3.85

0.56

0.87

NOTE: Inflation, π , and the nominal interest rate, R, are
expressed in percentage points.
SOURCE: Schmitt-Grohé and Uribe (2004b).

instrument. Specifically, it is assumed that the
government can levy distortionary income taxes
and can issue nominal non-state-contingent debt.
In financing innovations to its budget, the government can therefore either adjust distortionary income
taxes or adjust real public liabilities through an
appropriate price level change. In the theoretical
environment of Chari, Christiano, and Kehoe surprise changes in the price level are non-distorting,
whereas changes in the tax rates are distorting. As
a consequence, the optimal fiscal and monetary
policy mix calls for stable tax rates and highly volatile
inflation rates.
Clearly, this branch of the theoretical literature
is at odds with the goal of price stability. In the
mid-1990s, rigidities in product and factor price
adjustment found renewed attention in monetary
economics. Under sticky prices, both the predictions
about the optimal level and the volatility of inflation
may change. First, authors such as Goodfriend and
King (1997) showed that in simple models with price
stickiness but no money, the optimal inflation rate
is zero at all times and under all circumstances.
Khan, King, and Wolman (2003) show that if one is
to introduce money into the sticky-price model, a
tension arises between nominal interest rate stabilization at zero—to minimize the distortions associated with money—and the sticky-price friction,
which calls for constant prices at all times and under
all circumstances. Khan, King, and Wolman show
that for realistic calibrations of their model, the
optimal level of inflation is just –76 basis points.
44

J U LY / A U G U S T 2 0 0 4

(In their calibration, the Friedman rule would call
for inflation of –293 basis points and, absent the
monetary distortion, optimal policy is associated
with zero inflation.) That is, the optimal long-run
inflation rate is not that different from the one that
is optimal in a sticky-price model without money.
Schmitt-Grohé and Uribe (2004b) study optimal
monetary and fiscal policy in a model with (i) sticky
prices, (ii) money demand, (iii) distortionary taxation,
and (iv) a fiscal role for price level variations. In that
economy, as in the work of Chari, Christiano, and
Kehoe (1991), price level variations are desirable
because they allow the fiscal authority to finance
surprises in its budget by inflating or deflating the real
value of government debt rather than via changes in
distortionary income tax rates. Therefore, there exist
additional reasons to deviate from price stability
beyond those present in Khan, King, and Wolman.
Yet, for a model calibrated to the U.S. economy,
Schmitt-Grohé and Uribe (2004b) find that under the
Ramsey policy, the mean of inflation and its standard
deviation is close to zero. As shown in Table 1, in
the sticky-price economy, the Ramsey optimal mean
rate of inflation is only –0.16 percent and the optimal inflation volatility 0.17 percent. By contrast,
under fully flexible prices, the mean inflation rate is
–3.66 percent and the standard deviation of inflation
is 6.04 percentage points.
Furthermore, Schmitt-Grohé and Uribe (2004b)
show that the inflation-volatility tax-rate-volatility
trade-off is resolved in favor of inflation stability not
only for degrees of price stickiness observed in the
U.S. economy, but also for much lesser degrees of
price stickiness. This point is illustrated in Figure 1,
which shows on the horizontal axis the degree of
price stickiness as measured by a parameter θ. When
prices are perfectly flexible, then the parameter θ
is equal to zero. As the parameter θ increases, prices
become more sticky. The value of price stickiness
estimated for the U.S. economy is 4.4. The graph
shows that even for a degree of price stickiness ten
times smaller than the value estimated for the postwar U.S. economy, the optimal inflation volatility is
below 1 percent per year. These findings are further
support for Woodford’s conclusion that, in most
existing work, price stability should be the central
goal of optimal monetary policy.
This conclusion is based on evidence that relies
exclusively on models without an accumulable factor
of production. Next, I discuss some evidence on
the desirability of price stability in models where

FEDERAL R ESERVE BANK OF ST. LOUIS

physical capital is a factor of production and can be
accumulated. The basic elements of the model with
capital accumulation are, as in the one discussed
above, that money facilitates purchases of goods,
product markets are monopolistically competitive,
the government must finance a stochastic stream
of public consumption either with lump-sum or
income taxes, and prices are sticky à la Calvo (1983).
The production technology is described by some
homogenous-of-degree-one function and is subject
to shocks to total factor productivity (zt F(Kt, Ht )). The
evolution of capital is given by Kt+1=(1– δ ) Kt –1+It.
In Schmitt-Grohé and Uribe (2003), we compute
welfare in that economy under a number of alternative monetary and fiscal policy arrangements.
One of the policies considered is one in which the
inflation rate is held forever constant. We refer to
that policy as inflation targeting. We compute the
welfare consequences for the various rules under
the assumption that business cycles are driven by
government purchases and total factor productivity
shocks. We calibrate the model to the U.S. economy.
We consider monetary policy rules of the type
ˆ t = αRR
ˆ t −1 + α π πˆ t − j + α y ˆyt − j ,
(1) R

for j = −1, 0, +1,

where a hat over a variable indicates log-deviations
from its non-stochastic steady-state value and
(2)

ˆt = R
ˆ t −1 + α π πˆ t + α y ln yt / yt −1.
R

The variable measuring the output gap here is ŷt,
which denotes the deviations of output from the
non-stochastic steady state. Related studies typically
use a different measure of the output gap—namely,
one that measures the log-difference between the
actual level of output and the one that would obtain
in a model without price-adjustment frictions. (This
is what Woodford refers to as the properly defined
real stabilization objective.) Note that this is not
simply the difference of output from a linear time
trend, rather this is a highly sophisticated concept.
To be able to estimate that output gap, one needs
to know what the current realizations of the shocks
are and one has to know exactly where the nominal
frictions lie and how to compute the flexible-price
equilibrium.
The advantage of the interest rate feedback rule
given in equation (2) is that it puts even fewer informational requirements on the monetary authority.
To implement this rule, all the central bank needs to
know are the current values of output and inflation,
the past value of the nominal interest rate and output,

Schmitt-Grohé

Figure 1
Degree of Price Stickiness and Optimal
Inflation Volatility
Standard Deviation of π
7
6
5
4

← Baseline

3

← Flexible Prices

2
1
0
0

2

4

6

8

10

Degree of Price Stickiness, θ

NOTE: The baseline value of θ is 4.4. The standard deviation
is measured in percent per year.
SOURCE: Schmitt-Grohé and Uribe (2004b).

and the central bank’s inflation target, π*. The inflation target is needed to compute π̂t. Note that this rule
does not require knowledge of the non-stochastic
steady state; in particular, it is not necessary to know
the non-stochastic steady-state value of output or
the nominal interest rate.
For each case that we consider, we find that the
highest level of welfare is attained under a policy
of inflation targeting—that is, when the central bank
conducts policy in such a way that in equilibrium
the inflation rate is equal to its non-stochastic steadystate value at all times. This finding suggests that
even in models with capital, money, and distorting
taxes (but flexible wages), price stabilization should
be the overriding goal of policy. Table 2 illustrates
this point for the case that all taxes are lump-sum
and the economy is cashless. Similar results hold
for the monetary economy and in the presence of
distorting taxes. Inflation targeting yields at least as
much welfare as any of the optimized rules considered. Thus it provides further evidence that inflation
stability is desirable.
The table suggests two other interesting
results. One is that it is optimal not to respond to
output. This is reflected in the fact that the optimal
response coefficient on output is zero in almost all
cases. The second is that the welfare differences
J U LY / A U G U S T 2 0 0 4

45

REVIEW

Schmitt-Grohé

Table 1
Desirability of Price Stability in a Model with Capital Accumulation
απ

αy

αR

Welfare

Welfare cost

0.9

–628.2180

0

I. Monetary policy: R̂t = απ π̂t–i + α yŷt– i + α R R̂t–1
Smoothing
Current-looking (i = 0)

3

0

Backward-looking (i = 1)

3

0

2.8

–628.2207

0.0004

Forward-looking (i = –1)

3

0

–2.3

–628.8657

0.0886

Current-looking (i = 0)

3

0

—

–628.2193

0.0002

Backward-looking (i = 1)

3

–1.2

—

–629.2988

0.1477

No smoothing

Forward-looking (i = –1)

The equilibrium is indeterminate

II. Monetary policy: R̂t – R̂t–1 = απ π̂t + α y[ŷt – ŷt–1]
3

0

—

–628.2180

<0

–628.2175

–0.00007

III. Monetary policy: inflation targeting π̂t = 0

NOTE: (i) Rt denotes the gross nominal interest rate, πt denotes the gross inflation rate, and yt denotes output. (ii) For any variable xt ,
its non-stochastic steady-state value is denoted by x, and its log-deviation from steady state by x̂t ; ln(xt /x). (iii) In all cases, the parameters
απ , αy , and αR are restricted to lie in the interval [–3,3]. (iv) Welfare is defined as follows: Let V(st ) denote the equilibrium level of lifetime
utility of the representative household in period t given that period’s state st . Then welfare is defined as V(s). (v) The welfare cost is
relative to the optimal current-looking rule with smoothing and is defined as the percentage decrease in the consumption process
associated with the optimal rule necessary to make the level of welfare under the optimized rule identical to that under the alternative
policy considered. Thus, a positive figure indicates that welfare is higher under the optimized rule than under the alternative policy
considered. (vi) Computations are based on a second-order approximation.
SOURCE: Schmitt-Grohé and Uribe (2003).

between the various optimized rules and inflation
targeting are negligible from a welfare point of view
as long as the central bank has the option to smooth
interest rates over time.
The second point raises an interesting issue.
Clearly, one would like to know what the optimal
monetary policy is. But, at the same time, it is also
important to gauge how costly it would be to pursue
a policy that is not the optimal one but something
that one could realistically implement in practice. For
example, the optimal policy prescription presented
by Woodford in the appendix (drawing on Giannoni
and Woodford, forthcoming) is based on an estimated model with wage and price stickiness. Consequently, one would expect the optimal policy to be
quite complex, and indeed so it is. Its characterization
involves not only current values but also infinite-lead
polynomials of wages, prices, and a sophisticated
output gap measure. So it is natural to ask how much
of a quantitative difference it would make in terms
46

J U LY / A U G U S T 2 0 0 4

of welfare to follow this optimal policy prescription
as opposed to a much simpler one. In addition, it
would be useful to know whether it is important to
get the response coefficients exactly right or whether
there exists a large family of rules that are associated
with welfare levels that are very close to the level
of welfare associated with the optimum. These are
quantitative questions that are necessarily model
specific.
As a first pass on this question, I will present
some numerical results from the economy studied
in Schmitt-Grohé and Uribe (2003) described above.
In that particular framework there may be very little
difference between a large number of monetary
policies that the central bank can follow. The economy of Schmitt-Grohé and Uribe (2003) differs from
the Giannoni and Woodford economy in several
dimensions. Our model features capital accumulation, whereas the Giannoni and Woodford economy
does not. On the other hand, the Giannoni and

FEDERAL R ESERVE BANK OF ST. LOUIS

Schmitt-Grohé

Figure 2
Determinancy Regions and Welfare in the Model with
Capital
R̂t= α R R̂t–1 + απ π̂ t
(α y=0)
3

2

1

αR 0

–1

–2

–3
–3

NOTE:Abinatofrwh
ecmdps ichteqis ulbrm
d cirled
etrmina.A
optimzedrul(i.,
SOURCE:chmit-Gro

–2

–1

noteshawelfareco
α=π3,=0, αy
! andUribe(203).

0
απ

1

2

3

tiveohsfplcyra
and=0.9esthan0.5prc)il αR

Woodford framework features habit formation, sticky
wages (which is an important difference, as it makes
price stabilization less desirable), and some decision
lags that make that framework match estimated
empirical impulse responses. Ours is not an estimated model.
Figure 2 shows regions of the interest rate feedback rule coefficients αR and απ introduced in equation (1) for which the welfare cost of following that
policy (as opposed to the optimized rule) is at most
5 one-hundredths of 1 percent of the consumption
stream associated with the optimized rule. In these
computations the output response coefficient is held
constant at zero (α y=0). The graph shows that the
region of parameters for which the equilibrium is
unique is virtually the same as the region for which
the welfare costs are below 5 basis points. This suggests that from a welfare point of view, it does not
really matter to which values the central banks sets

the response coefficients in the interest rate feedback
rule as long as they render the equilibrium determinate. Similar results hold for the feedback rule
given in equation (2), with α y equal to its optimized
value of zero (see Schmitt-Grohé and Uribe, 2003).
My last comment concerns the implementation
of inflation targeting. In particular, one interesting
issue is whether the emphasis on price stability that
comes out of the theoretical literature on optimal
monetary policy implies that interest rates should
respond little to output measures. Suppose the central bank chooses to implement its policy objectives
by following a feedback rule for the short-term
nominal interest rate that it controls. The results
presented in Table 2 indicate that the optimal
response coefficient on the output gap should be
zero, where the output gap is defined as the logdeviation of output from steady state. As we show
in Schmitt-Grohé and Uribe (2003), the welfare losses
J U LY / A U G U S T 2 0 0 4

47

REVIEW

Schmitt-Grohé

from choosing a non-zero coefficient on output can
be large. We find values in excess of one-tenth of 1
percent of the stream of consumption associated
with the optimized rule.1 The argument that responding to output can lead to relatively sizeable welfare
losses has been criticized on the grounds that the
output measure used in the monetary policy rule is
“not the right one.” The argument goes that, were
one instead to use an output gap measure based on
the difference between actual output and the output
that would arise in a world without nominal frictions,
then the welfare losses from responding to output
in the feedback rule would be much smaller.
In practice the central bank may not be able to
construct this sophisticated output gap measure
and will instead use a simple measure that is much
more akin to log deviations from a constant trend.
Furthermore, one can show that if the output gap is
interpreted as the difference of the quarterly output
growth rate from some constant, then welfare losses
associated with responding to that measure of output are small as well. Under such a rule it is still
optimal not to respond to output (see the second
panel of Table 2). However, Schmitt-Grohé and Uribe
(2003) show that the welfare differences between a
zero output coefficient and an output coefficient
between –3 and 3 is at most 0.03 percentage points
of the consumption stream associated with the best
feedback rule. This welfare loss is relatively small.
These findings suggest that when implementing
inflation targeting through interest rate feedback
rules it may suffice to respond to variations in inflation alone. Second, a reason for policymakers to
abstain from responding to output variations is that
such behavior may have significant welfare consequences if the policymaker does not have the proper
output gap measure. It is important to keep in mind
that the optimal policy behavior advocated here does
not have stabilization of inflation as its ultimate
objective, but instead the maximization of welfare.
Thus, even though the implementation of the optimal
policy takes the form of a rule that responds little
1

In a simple model in which complete inflation stabilization is optimal,
Rotemberg and Woodford (1997, Table 2) report the value of a loss
function that is a linear transformation of the unconditional expectation
of the utility function for various values of the feedback rule coefficients
απ and αy. For example, for values of απ=1.5 and αy=0.5, the loss function is 8.72, whereas for απ=10 and α y=0, it is only 0.93. These numbers suggest that not responding to output is desirable and that higher
output response coefficients are associated with lower unconditional
welfare. However, from those numbers one cannot tell whether the
welfare losses would be large or small in terms of units of consumption.

48

J U LY / A U G U S T 2 0 0 4

to variations in the level of aggregate activity, this
does not imply that the reasons for adopting this
policy are that the policymaker does not fully internalize the welfare consequences of output fluctuations. One caveat is that these recommendations
stem from an analysis in which factor prices are
assumed to be fully flexible. It remains to be shown
in future work how large the welfare costs or benefits
of responding to output are in a world with sluggish
factor price adjustments.

REFERENCES
Calvo, Guillermo A. “Staggered Prices in a Utility-Maximizing
Framework.” Journal of Monetary Economics, 1983,
12(3), pp. 383-98.
Chari, Varadarajan V.; Christiano, Lawrence and Kehoe,
Patrick J. “Optimal Fiscal and Monetary Policy: Some
Recent Results.” Journal of Money, Credit, and Banking,
August 1991, 23(3, Part 2), pp. 519-39.
Giannoni, Marc P. and Woodford, Michael. “Optimal
Inflation Targeting Rules,” in Ben S. Bernanke and
Michael Woodford, eds., The Inflation Targeting Debate.
Chicago: University of Chicago Press (forthcoming).
Goodfriend, Marvin and King, Robert G. The New
Neoclassical Synthesis and the Role of Monetary Policy,”
in Ben S. Bernanke and Julio J. Rotemberg, eds., NBER
Macroeconomics Annual 1997. Cambridge, MA: MIT Press,
1997, pp. 231-83.
Khan, Aubhik; King, Robert G. and Wolman, Alexander.
“Optimal Monetary Policy.” Review of Economic Studies,
October 2003, 70, pp. 825-60.
Rotemberg, Julio J. and Woodford, Michael. “An OptimizationBased Econometric Framework for the Evaluation of
Monetary Policy,” in Ben S. Bernanke and Julio J.
Rotemberg, eds., NBER Macroeconomics Annual 1997.
Cambridge, MA: MIT Press, 1997, pp. 297-346.
Schmitt-Grohé, Stephanie and Uribe, Martín. “Simple
Optimal Implementable Monetary and Fiscal Policy.”
Unpublished manuscript, Duke University, October 2003.
Schmitt-Grohé, Stephanie and Uribe, Martín. “Solving
Dynamic General Equilibrium Models Using a SecondOrder Approximation to the Policy Function.” Journal of
Economic Dynamics and Control, January 2004a, 28(4),
pp. 755-75.

FEDERAL R ESERVE BANK OF ST. LOUIS

Schmitt-Grohé

Schmitt-Grohé, Stephanie and Uribe, Martín. “Optimal
Fiscal and Monetary Policy Under Sticky Prices.” Journal
of Economic Theory, February 2004b, 114(2), pp. 198-230.
Sims, Christopher A. “Notes on the GYNSYS2 Packet.”
Unpublished manuscript, Princeton University, 2000.

J U LY / A U G U S T 2 0 0 4

49

REVIEW

50

J U LY / A U G U S T 2 0 0 4

The Macroeconomic Effects of Inflation Targeting
Andrew T. Levin, Fabio M. Natalucci, and Jeremy M. Piger

1. INTRODUCTION

O

ver the past 15 years, explicit inflation
targeting (IT) has been adopted by an
increasing number of central banks, and
a substantial body of literature has emphasized the
advantages of this approach as a framework for
monetary policy.1 Nevertheless, empirical analysis
has yielded little evidence of any macroeconomic
effects of IT. For example, the landmark study of
Bernanke, Laubach, Mishkin, and Posen (1999) concluded that the first few countries to adopt IT did
not experience any short-run gains in lower output
costs of disinflation. Most recently, Ball and Sheridan
(forthcoming) considered a wide range of macroeconomic indicators for Organisation for Economic
Cooperation and Development (OECD) economies
and found no statistically significant differences
between the IT and non-IT countries.
In this paper, we evaluate the extent to which
IT exerts a measurable influence on expectations
formation and inflation dynamics. For the industrialized economies, we address this question by comparing time-series data since 1994 for five IT countries
(Australia, Canada, New Zealand, Sweden, and the
United Kingdom) with that of seven non-IT countries
(the United States, Japan, Denmark, and four of the
five largest euro area members—namely, France,
Germany, Italy, and the Netherlands).2 For these
1

See Leiderman and Svensson (1995), Bernanke and Mishkin (1997),
Bernanke et al. (1999), Schaechter, Stone, and Zelmer (2000), Corbo,
Landerretche, and Schmidt-Hebbel (2001), Mishkin and Schmidt-Hebbel
(2001), Neumann and von Hagen (2002), Benati (2003), Goodfriend
(forthcoming), and Svensson and Woodford (forthcoming).

2

To avoid consideration of structural breaks midway through the sample,

economies, we analyze the behavior of medium- and
long-term inflation expectations using Consensus
Economics Inc. semiannual surveys of market forecasters, and we employ the methods of Stock (1991)
and Hansen (1999) to obtain median-unbiased measures of persistence for total and core consumer price
inflation (CPI). Finally, since the experience with IT
in the emerging market economies (EMEs) is mainly
limited to the past few years, our analysis of these
economies follows an event-study approach similar
to that of Bernanke et al. (1999).
For the industrialized economies, our evidence
indicates that IT has played a significant role in
anchoring long-run inflation expectations. For the
United States and the euro area, private-sector inflation forecasts (at horizons up to ten years) exhibit a
highly significant correlation with a three-year moving average of lagged inflation.3 In contrast, at the
longest horizons this correlation is largely absent
for the five IT countries, indicating that these countries’ central banks have been quite successful in
delinking expectations from realized inflation.4
We also find that actual inflation exhibits markour analysis excludes Norway and Switzerland (which adopted explicit
inflation targets in 2000 and 2001, respectively) as well as Finland
and Spain (which moved from IT to euro area membership). See
Dueker and Fisher (1996).
3

In related work, Gurkaynak, Sack, and Swanson (2003) find evidence
that shifts in private-market perceptions about long-term inflation
account for a substantial proportion of the degree to which U.S. longterm bond rates are highly sensitive to federal funds rate surprises.
See also Bernanke and Kuttner (2003), Bonfim (2003), and Kozicki
and Tinsley (2001a,b).

4

For results regarding the effects of IT on short-term inflation expectations, see Johnson (2002, 2003) and Gavin (2004).

Andrew T. Levin is a senior economist in the division of monetary affairs and Fabio M. Natalucci is an economist in the division of international
finance at the Board of Governors of the Federal Reserve System. Jeremy M. Piger is an economist at the Federal Reserve Bank of St. Louis. The
authors appreciate comments from Guy Debelle, Bill Gavin, Ken Kuttner, Athanasios Orphanides, Bob Rasche, Dan Thornton, Harald Uhlig, and
seminar participants at the Federal Reserve Bank of St. Louis 28th Annual Economic Policy Conference, “Inflation Targeting: Prospects and Problems.”
Jane Ihrig graciously provided a summary of institutional features of inflation targeting in the industrial economies. Claire Hausman, Michelle Meisch,
Ryan Michaels, and Clair Null provided excellent research assistance. The views expressed in this paper are solely the responsibility of the authors
and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System, the Federal Reserve Bank of St. Louis,
or of any other person associated with the Federal Reserve System.
Federal Reserve Bank of St. Louis Review, July/August 2004, 86(4), pp. 51-80.
© 2004, The Federal Reserve Bank of St. Louis.

J U LY / A U G U S T 2 0 0 4

51

Levin, Natalucci, Piger

edly lower persistence in IT countries.5 For example,
even with only a decade of quarterly data, we can
clearly reject the null hypothesis of a unit root in
core CPI inflation for Canada, New Zealand, Sweden,
and the United Kingdom. Inflation persistence is
estimated to be quite low in these countries, with the
90 percent confidence interval for the largest autoregressive root excluding 0.7 in all cases. By contrast,
the unit-root null hypothesis cannot be rejected for
the United States, the euro area, or Japan.6
For the EMEs, the initial experience with IT
appears to be largely consistent with that observed
by Bernanke et al. (1999) for the industrialized countries.7 In particular, our event-study approach confirms that the adoption of IT is not associated with
an instantaneous fall in private-sector inflation forecasts, especially at longer horizons. Since measures
of potential output and the natural unemployment
rate are notoriously difficult to construct for EMES,
we have not attempted to compute sacrifice ratios
for these episodes; however, informal assessment
suggests that the adoption of IT was not associated
with a marked reduction in the output costs of
disinflation.
It should be noted that the absence of instantaneous gains from IT is not necessarily inconsistent
with substantial macroeconomic effects over a
period of a decade or more. If an economy has
already been experiencing low and stable inflation
for an extended period, then the adoption of a formal
IT regime might not have any immediate benefit—
the delinking of expectations from realized inflation
would only become visible at some later date when
the economy was hit by a substantial shock. On the
other hand, if IT is adopted at a point of relatively
high or volatile inflation, then the private sector
5

Siklos (1999) finds evidence of a decline in inflation persistence in
some IT countries; see also Kuttner and Posen (1999). Using a sample
of more than 100 countries, Kuttner and Posen (2001) find evidence
that inflation-targeting countries experience lower inflation persistence.
Corbo et al. (2001) find that IT is associated with lower long-term
effects of inflation innovations compared with the non-IT countries.

6

As shown in section 3, we find that the unit root null hypothesis can
be rejected for U.S. total CPI inflation but not for core CPI inflation. A
number of studies have considered the extent to which recent U.S.
inflation data exhibits less persistence than that of a random walk;
cf. Barsky (1987), Evans and Wachtel (1993), Fuhrer and Moore (1995),
Brainard and Perry (2000), Taylor (2000), Cogley and Sargent (2002,
2003), Kim et al. (2001), Stock (2002), Pivetta and Reis (2001), Levin
and Piger (2002), and Benati (2002). For estimates of inflation persistence for other countries, see also Ravenna (2000) and Batini (2002).

7

See also Ammer and Freeman (1995), Laubach and Posen (1997),
Almeida and Goodhart (1998), and Corbo et al. (2001).

52

J U LY / A U G U S T 2 0 0 4

REVIEW
might reasonably be skeptical about the likely duration of the regime, and hence its inflation expectations would only adjust gradually (cf. Erceg and
Levin, 2003).
Finally, our analysis underscores the key role of
institutional considerations in determining inflation
expectations. In particular, as emphasized by Kohn
(forthcoming), the volatility of long-term inflation
expectations for a number of IT countries is roughly
similar to that of some non-targeters such as the
United States. Since our analysis suggests that the
IT countries have succeeded in delinking inflation
expectations from lagged inflation, the ongoing
fluctuations in long-term expected inflation for
these countries are evidently related to shifting views
about the long-term course of monetary policy (e.g.,
the probability that Sweden or the United Kingdom
might join the European Monetary Union).
The remainder of this paper is organized as follows. For the industrial economies, section 2 presents
our findings on the determination of inflation expectations, section 3 reports our results regarding inflation persistence, and section 4 presents evidence
regarding macroeconomic volatility. For the EMEs,
section 5 provides an overview of IT arrangements,
and section 6 presents our event-study analysis of
the initial effects of IT. Section 7 summarizes our
conclusions and discusses some areas for future
research.

2. INFLATION TARGETING AND
INFLATION EXPECTATIONS IN
INDUSTRIALIZED ECONOMIES
In this section we begin our analysis of the
macroeconomic effects of IT by investigating the
behavior of inflation expectations in our sample of
IT and non-IT economies. We are primarily interested
in whether inflation expectations, particularly at
longer horizons, are relatively more anchored in IT
economies.
To measure inflation expectations, we use survey
results collected by Consensus Economics. Twice
each year, market forecasters are polled regarding
their inflation forecasts at horizons of one to ten
years. The mean panelist forecast serves as our
measure of inflation expectations. We obtained
these forecasts from 1994 to the present for each
of the countries in our samples, with the exception
of Denmark. In the results presented here, the “euro
average” we form is a weighted average of France,
Germany, Italy, and the Netherlands using GDP shares

FEDERAL R ESERVE BANK OF ST. LOUIS

as weights. Thus, our non-IT sample consists of
this euro average, Japan, and the United States.
Figure 1 displays the inflation expectation series
for four forecast horizons: one, three, five, and sixto-ten years ahead. Note that in many cases, the
series drift downward over the early part of the
sample. To account for this nonstationarity, the
empirical results presented in subsequent sections
focus on first differences of the expectation series.

Levin, Natalucci, Piger

Table 1
Standard Deviation of Change in Inflation
Expectations (1994-2003)
Horizon (years ahead)
1

2.2 Sensitivity of Expectations to
Realized Inflation
We now estimate the sensitivity of inflation
expectations to realized inflation in IT and non-IT
countries. In particular, we estimate a pooled regression in which the left-hand-side variable is the first

5

6-10

IT sample

2.1 Volatility of Inflation Expectations
As a first pass at investigating these data, Table 1
presents the standard deviation of the first difference
of the expectations series for the four forecast horizons plotted in Figure 1.
Overall, the results in Table 1 suggest that inflation expectations are not noticeably more volatile
in non-IT vs. IT economies. Indeed, expectations
for the euro average and the United States are less
volatile than the average for the IT economies at
every forecast horizon and display similar or less
volatility than most of the individual IT economies.
On the other hand, Japanese inflation expectations
are much more volatile than the other economies,
particularly at longer horizons.
These results are consistent with those of Kohn
(forthcoming), who used Consensus Economics’
measures of inflation expectations and found that
the volatility of changes in inflation expectations in
Germany and the United States are no higher than
those in Canada, Sweden, and the United Kingdom.
Nevertheless, even if the unconditional volatility of
inflation expectations is no less in IT economies,
expectations may still be more anchored in IT
economies in that they are less responsive to macroeconomic developments. That is, two countries with
identical inflation expectation volatility may have
such volatility for very different reasons. For example,
suppose that IT has anchored inflation expectations
in the United Kingdom, making them less responsive
to macroeconomic fluctuations. In this case, inflation
expectations may still be unconditionally relatively
volatile, due to, say, institutional uncertainty surrounding the possible adoption of the euro.

3

Australia

0.76

0.36

0.41

0.16

Canada

0.33

0.23

0.17

0.21

New Zealand

0.53

0.19

0.16

0.13

Sweden

0.44

0.24

0.19

0.26

United Kingdom

0.16

0.17

0.17

0.21

IT mean

0.44

0.24

0.22

0.19

Euro average

0.22

0.14

0.15

0.10

Japan

0.42

0.40

0.39

0.66

United States

0.25

0.21

0.16

0.11

Non-IT sample

NOTE: This table contains the standard deviation of the first
difference of the mean inflation forecast collected by Consensus
Economics Inc. over the period 1994 through the second half
of 2003. The “euro average” is a weighted average of France,
Germany, Italy, and the Netherlands, using GDP shares as
weights.

difference of inflation expectations and the righthand-side variable is the first difference of lagged
realized CPI inflation. Formally, we estimate the
following equation:
(1)

∆π̂ i(,qt ) = λ i + β ∆ π i ,t + ε i ,t ,

where π̂ i(,qt ) is an expectation of inflation q years in
the future in country i, formed at time t, and π–i,t is
a three-year moving average of inflation in country
i ending at time t. Equation (1) is estimated for our
sample of both IT economies and non-IT economies,
yielding an estimate of β for each set of countries.
Given the relatively high level of expectations volatility in Japan, and the fact that economic performance
in Japan has been quite different from that in the euro
area and the United States over this sample period,
we also present estimates for a non-IT sample consisting of the euro average and the United States only.
Table 2 reports estimates of the relationship
between realized inflation and expected inflation
at several different forecast horizons. These estimates
suggest that longer-run inflation expectations have
been much less responsive to actual inflation develJ U LY / A U G U S T 2 0 0 4

53

REVIEW

Levin, Natalucci, Piger

Figure 1

Inflation Expectations
One Year Ahead
4.5

4

4.0
3
3.5
3.0

2

2.5
1

2.0
1.5

0
1.0
0.5
1994

1996

1998

2000

–1
1994

2002

1996

1998

1996

1998

2000

2002

Three Years Ahead
4

4.0
3.5

3

3.0
2
2.5
1
2.0
0

1.5

1.0
1994

1996

1998

Australia
Canada
New Zealand

2000

2002

Sweden
United Kingdom

opments in IT countries than in non-IT countries.
At the five-year horizon, the estimated response of
the change in expected inflation to the change in
lagged actual inflation in non-IT economies is over
three times that in IT economies. At the six-to-tenyear horizon, the estimated response in non-IT
economies is still around 25 basis points, whereas
the estimated response in IT countries is close to
zero and statistically insignificant. This suggests
that IT central banks have been quite successful in
54

J U LY / A U G U S T 2 0 0 4

–1
1994

Euro Average

2000
Japan

2002
United States

delinking expectations from realized inflation. The
final row of the table demonstrates that these results
are robust to removing Japan from the non-IT group.
Some have argued that, in the United States, the
Federal Reserve pursued a policy of “opportunistic
disinflation” during the early years of our sample
and that the dynamics of inflation are likely different
in the years following this disinflation. To investigate
this possibility, we estimated equation (1) for U.S.
data only, over a sample beginning in 1998 rather

FEDERAL R ESERVE BANK OF ST. LOUIS

Levin, Natalucci, Piger

Figure 1 cont’d

Inflation Expectations
Five Years Ahead
4.0

4.0
3.5

3.5

3.0
3.0

2.5

2.5

2.0
1.5

2.0

1.0
1.5

0.5

1.0
1994

1996

1998

2000

2002

0.0
1994

1996

1998

2000

2002

1996

1998

2000

2002

Six-to-Ten Years Ahead
4.0

4

3.5
3
3.0

2.5

2

2.0
1
1.5

1.0
1994

0
1996

1998

Australia
Canada
New Zealand

2000

2002

Sweden
United Kingdom

1994

Euro Average

Japan

United States

NOTE: This figure contains the mean inflation forecast collected by Consensus Economics Inc. over the period 1994 through
2003. "Euro average" is a weighted average of France, Germany, Italy, and the Netherlands, using GDP shares as weights.
For Japanese five-year-ahead and six-to-ten year-ahead expectations, the observation for the second half of 1997 was
missing. This was replaced by the median of the six adjacent observations for this figure.

than in 1994. Over this period, the estimated
response of five-year-ahead inflation expectations
on lagged three-year average inflation is 0.34, with
a standard error of 0.13, which is similar to the
estimate of 0.36 obtained for the U.S. data over the
sample beginning in 1994. Also, it is interesting to
note that the most recently obtained observation

for five-year-ahead U.S. inflation expectations,
released by Consensus Forecasts in mid-October
of 2003, declined from 2.5 to 2.2 percent, which
corresponds to a decline in lagged three-year average
inflation from 2.5 to 2.3 percent.
These findings are broadly consistent with those
reported by Castelnuovo, Nicoletti-Altimari, and
J U LY / A U G U S T 2 0 0 4

55

REVIEW

Levin, Natalucci, Piger

Table 2
Estimated Response of Change in Inflation
Expectations to Change in Realized Inflation
Horizon (years ahead)
1

3

5

6-10

IT

0.00
(0.10)

0.20
(0.06)

0.09
(0.05)

0.01
(0.05)

Non-IT

–0.03
(0.17)

0.25
(0.11)

0.29
(0.11)

0.24
(0.08)

Euro area and
United States

–0.06
(0.19)

0.30
(0.12)

0.34
(0.11)

0.24
(0.08)

NOTE: This table holds estimates of β from equations (1) and
(2) applied to both IT and non-IT economies over the period
1994-2003. Standard errors are in parentheses. Estimation was
performed via generalized least squares assuming cross-sectional
heteroskedasticity. Similar results are obtained when estimation
is performed via a seemingly unrelated regression.

Palenzuela (2003; CNP), who analyzed the relationship between changes in long-term expected inflation (at a horizon of six-to-ten years) and changes in
one-year-ahead expected inflation. Using Consensus
Economics’ survey data for the period 1995-2002,
CNP obtained regression coefficients of 0.21 for the
United States, 0.31 for Switzerland, and 0.43 for
Japan, compared with an average coefficient of 0.13
for the five IT countries in our sample. For the euro
area, CNP used the sample period 1999-2002 and
obtained a regression coefficient of 0.08, closer to
that of IT economies than non-IT economies.
Finally, it is interesting to note that Ball and
Sheridan (forthcoming) found that one-year-ahead
inflation expectations are about one-third less
responsive to realized inflation developments in IT
economies than in non-IT economies, but this difference is not statistically significant. This evidence is
consistent with the results in Table 2 for forecast
horizons of one and three years. However, it seems
reasonable that IT, by revealing a long-run trend
rate of inflation, would have its greatest chance of
success at anchoring long-horizon expectations.
Indeed, the results in Table 2 suggest that longrun inflation expectations are substantially more
anchored in IT economies.8

3. INFLATION TARGETING AND
INFLATION DYNAMICS IN
INDUSTRIALIZED ECONOMIES
In the previous section, we studied the behavior
of inflation expectations in IT and non-IT economies.
In this section we turn our analysis to the dynamics
of actual inflation. We are particularly interested in
whether inflation persistence is lower in IT countries
than in non-IT countries.

3.1 A Look at the Data
Our data consist of inflation rates for our sample
of IT economies (Australia, Canada, New Zealand,
Sweden, and the United Kingdom) and non-IT
economies (Denmark, France, Germany, Italy, the
Netherlands, Japan, and the United States). We also
consider a euro-area average inflation rate, which
is average inflation across the 12 countries that have
adopted the euro; the sample period runs from the
first quarter of 1994 to the second quarter of 2003
for all countries.
For each country, we analyze two measures of
inflation, the first based on the total CPI and the
second based on the core CPI, measured as the total
CPI less food and energy prices. Inflation is calculated
as the annualized quarterly percentage change in
the price index. All data were obtained from the
OECD. We identify three specific cases in which
exogenous shifts in tax rates resulted in large transitory fluctuations in the inflation series. These consist
of the introduction of the goods and services tax
(GST) in Australia in the third quarter of 2000; large
changes in cigarette taxes in Canada in the first two
quarters of 1994; and an increase in the consumption tax in Japan in the second quarter of 1997. As
shown by Franses and Haldrup (1994), such outliers
can induce substantial downward bias in the estimated degree of persistence. Thus, before analyzing
the inflation series, we replace the outliers with
interpolated values (the median of the six adjacent
observations that were not themselves outlier observations). The total and core CPI inflation rates for
each country are shown in Figure 2.

3.2 Methodology
To measure inflation persistence, we estimate a
univariate autoregressive process for each inflation
series:
K

8

For more discussion of Ball and Sheridan (forthcoming), see Gertler
(forthcoming).

56

J U LY / A U G U S T 2 0 0 4

(2)

π t = µ + ∑ α jπ t − j + εt ,
j =1

FEDERAL R ESERVE BANK OF ST. LOUIS

Levin, Natalucci, Piger

Figure 2
Inflation Rates
7

6
6
4
4
3

3

2
2

1

1

0

0

–1

–1

–2
–3

–2
4

6

7

8

00

01

02

4

6

7

8

00

01

02

8

00

01

02

00

01

02

F
D

4

4

3
3
2
2

1
0

1

–1
0

–2
–3

–1
4

6

7

8

G

4

00

01

4

02

y

6

7

I

8

y

7
3
6
2
4
1
3
2

0

1
–1

0
4

6

7

8

00

01

02
I

4

6

7

8

I

J U LY / A U G U S T 2 0 0 4

57

REVIEW

Levin, Natalucci, Piger

Figure 2 cont’d
Inflation Rates
Japan

4

Netherlands

6

3

5

2

4

1

3

0

2

–1

1

–2

0

–3
-3

-1
–1
94

95

96

97

98

99

00

01

02

94

New Zealand

6

95

96

97

98

99

00

01

02

00

01

02

00

01

02

Sweden

8
6

4
4
2

2
0

0

–2
–2
–4
–6
-6

–4
-4
94

95

96

97

98

99

00

01

94

02

United Kingdom

6

95

96

97

98

99

United States

4

5
3
4
2

3
2

1

1
0
0
–1
-1

–1
-1
94

95

96

97

98

99

00

01

02

94
CPI

58

J U LY / A U G U S T 2 0 0 4

Core CPI

95

96

97

98

99

FEDERAL R ESERVE BANK OF ST. LOUIS

Levin, Natalucci, Piger

Table 3
Persistence Estimates for Inflation
Core CPI
Country

Median unbiased

Total CPI

Upper 95th percentile

Median unbiased

Upper 95th percentile

IT countries
Australia

0.70

1.02

0.47

0.80

Canada

0.27

0.63

–0.22

0.21

New Zealand

0.24

0.60

0.25

0.61

Sweden

0.16

0.54

0.04

0.44

United Kingdom

0.33

0.68

0.06

0.45

Denmark

0.66

1.00

–0.74

–0.23

Euro area

0.84

1.06

0.87

1.06

France

0.75

1.04

0.91

1.07

Germany

0.77

1.04

0.81

1.05

Italy

0.88

1.07

0.88

1.07

Netherlands

0.39

0.74

0.51

0.83

Japan

0.82

1.05

0.72

1.03

United States

1.04

1.10

0.54

0.86

Non-IT countries

NOTE: For each country in the sample, this table records the median unbiased estimate and the upper bound of the two-sided 90
percent confidence interval for the largest autoregressive root of core and total CPI inflation, estimated over 1994:Q1–2003:Q2.
Estimates were computed based on Stock (1991), using equation (2).

where εt is a serially uncorrelated, homoskedastic
random error term. To obtain a scalar measure of
persistence from equation (2), we use the largest
autoregressive root, denoted ρ and defined as the
largest root of the characteristic equation
K

λK − ∑ α j λK − j = 0 . The largest autoregressive root
j =1

has intuitive appeal as a measure of persistence,
as it determines the size of the impulse response,
∂π t + j
, as j grows large. We apply the procedures
∂ε t
developed in Stock (1991) to obtain median unbiased
estimates and an upper 95th percentile estimate,
which is the upper bound of a two-sided 90 percent
confidence interval.
As a robustness check, we also consider an
alternative measure of persistence, namely, the sum
K

of the autoregressive coefficients, α ; ∑ α j . As
j =1

noted by Andrews and Chen (1994), α also has
intuitive appeal as a measure of persistence, as it is
monotonically related to the cumulative impulse

response of πt+j to εt. We construct a median unbiased and upper 95th percentile estimates for α using
the “grid bootstrap” procedure of Hansen (1999).
This technique simulates the sampling distribution
αˆ − α
of the t-statistic t =
over a grid of possible
se (αˆ )
true values for α to construct confidence intervals
with correct coverage.
To estimate (1), an autoregressive lag order K
must be chosen for each inflation series. For this
purpose, we utilize Akaike information criterion,
the information criterion proposed by Akaike (1973),
with a maximum lag order of K=4 considered. The
lag order chosen for each series is reported in
Appendix Table A1.

3.3 Persistence Estimates
We begin by discussing persistence estimates
for the core CPI. Table 3 presents these results for
each country in the sample. Note that values less
than unity for the upper 95th percentile estimate
imply that a unit root can be rejected for this series
J U LY / A U G U S T 2 0 0 4

59

REVIEW

Levin, Natalucci, Piger

Figure 3
Average Impulse Response Functions Based on Core CPI
1.2
IT
Non-IT
1

0.8

0.6

0.4

0.2

0
1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

Quarters

at the 5 percent level of significance (based on a
one-tailed test).
Consider first the results for the non-IT economies. Table 3 demonstrates that for Denmark, the
euro area, Japan, and the United States, the upper
95th percentile estimate is above unity, suggesting
that core CPI inflation in these economies displays
behavior consistent with a unit-root process. The
median unbiased estimate is also quite high in
general, above 0.8 for the euro area, Japan, and the
United States.
The results for the IT countries stand in contrast
to those for the non-IT economies. For Canada, New
Zealand, Sweden, and the United Kingdom, the
upper 95th percentile estimate is less than unity,
meaning that the unit root null hypothesis can be
rejected for these series. This is true even though the
sample size of roughly 40 observations is relatively
short. Indeed, the median unbiased estimate is
roughly 0.3 or less for these countries, which suggests a white noise process for inflation.
We now turn to the results for total CPI inflation,
also shown in Table 3. In this case, the evidence is
more mixed for the non-IT economies. In particular,
while the unit-root null hypothesis cannot be
60

J U LY / A U G U S T 2 0 0 4

rejected for both the euro area and Japanese inflation
rates, it is rejected for both Denmark and the United
States. For the IT economies, inflation persistence
is again estimated to be quite low, with the unit-root
null hypothesis rejected for all five IT countries.
Australia displays the highest median unbiased
point estimate of approximately 0.5 (similar to the
estimate for the United States), while the remaining
four countries have median unbiased estimates of
less than 0.3.9

3.4 Impulse Response Functions
An intuitive way to interpret our measures of
inflation persistence is to compute an impulse
response function, which gives the response of
inflation at various future dates to a shock that
occurs today. Figure 3 displays average impulse
response functions based on core CPI inflation both
for the five IT countries in our sample and for a nonIT sample consisting of Denmark, the euro area,
9

Note that estimated Australian inflation persistence is high relative to
the other IT economies for both core and total CPI inflation. We have
also estimated inflation persistence for the Australian CPI excluding
mortgage interest and obtained similar results to those for the total
CPI.

FEDERAL R ESERVE BANK OF ST. LOUIS

Levin, Natalucci, Piger

Figure 4
Average Impulse Response Functions Based on Total CPI
1.2
IT
Non-IT
Non-IT (excluding Denmark)
1

0.8

0.6

0.4

0.2

0
1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

Quarters

Japan, and the United States.10 The figure makes
clear that inflation shocks are much less persistent
in the sample of IT economies. For example, nearly
half of a one-unit shock to inflation in the IT economies has worn off after just one quarter, and 90
percent after just four quarters. By contrast, for the
non-IT sample, it is four quarters before half of the
effect of a one-unit shock has dissipated and eleven
quarters before this effect has fallen by 90 percent.
Impulse response functions can also help in
understanding the relationship between our results
and those of Ball and Sheridan (forthcoming), who
found no significant difference in the persistence of
total CPI inflation for IT and non-IT industrial economies. Figure 4 gives the average impulse response
functions for total CPI inflation. Consistent with Ball
10

Following Ball and Sheridan (forthcoming), the average impulse
response functions are computed by first averaging the autoregressive
(AR) coefficients across groups of countries and then computing an
impulse response function based on these average coefficients. For
simplicity, the impulse response functions are calculated based on an
AR(1) representation for inflation, with the AR(1) coefficients taken from
the median unbiased estimates for α reported in Appendix Table A2.
Thus, the impulse response functions are a smoothed version of an
impulse response function based on the full set of autoregressive
coefficients.

and Sheridan, the impulse response functions for
IT and non-IT economies are nearly identical, suggesting there are less-obvious differences in persistence between IT and non-IT economies.
The results for total CPI inflation are influenced
by the averaging of persistence estimates across
countries for the purpose of computing the impulse
response functions, which masks important details
about individual countries. For example, Denmark
displays considerable negative serial correlation for
total CPI inflation, which lowers the average impulse
response function for non-IT economies. This can
be seen in Figure 4, which also plots an average
impulse response function for the non-IT group,
excluding Denmark, and suggests greater differences
in persistence between the IT and non-IT group.

4. MACROECONOMIC VOLATILITY IN
INDUSTRIAL ECONOMIES
4.1 Output Volatility
One potential explanation for their damped
levels of inflation persistence is that IT countries
have practiced an active monetary policy, quickly
stamping out deviations of inflation from target levels.
J U LY / A U G U S T 2 0 0 4

61

REVIEW

Levin, Natalucci, Piger

Table 4
Standard Deviation of Core CPI Inflation and Real GDP Growth (1994-2003)
VAR(π t)/VAR(ε t)

Standard deviation, output

Standard deviation, inflation

Australia

2.54

1.73

2.04

Canada

2.03

0.93

1.23

New Zealand

3.97

2.10

1.16

Sweden

3.29

1.58

1.19

IT countries

United Kingdom

1.33

1.37

1.24

IT mean

2.63

1.54

1.37

3.34

0.90

1.07

Euro area

2.01

0.68

2.39

France

2.42

0.75

2.08

Germany

2.28

0.87

1.36

Italy

1.80

1.14

3.41

Non-IT countries
Denmark

4.34

0.90

1.29

Japan

Netherlands

1.46

0.75

1.73

United States

2.17

0.50

2.25

If this were the case, one would expect to see heightened levels of output volatility in IT countries where
the monetary authority manipulated the output gap
to reverse shocks to inflation (cf. Cecchetti and
Ehrmann, 1999). To investigate this potential explanation, the first column of Table 4 also reports the
standard deviation of real gross domestic product
(GDP) growth computed from 1994 to the present
for our sample of IT and non-IT economies.
As is apparent from the table, IT economies do
not seem to display heightened volatility of real GDP
growth relative to non-IT economies. In particular,
the five IT economies are spread relatively evenly
throughout the distribution of GDP volatility. This
suggests that the low levels of inflation persistence
in IT countries have not come at the expense of
heightened output-growth volatility. This suggests
that IT has improved the tradeoffs policymakers
face in these countries.11

4.2 Inflation Volatility: Propagation or
Shocks?
All else being equal, the relatively low levels of
inflation persistence documented for IT countries
11

For evidence regarding changes in output volatility across countries see
van Dijk, Osborn, and Sensier (2002) and Stock and Watson (forthcoming).

62

J U LY / A U G U S T 2 0 0 4

should suggest relatively low levels of unconditional
inflation volatility in these countries. However, as
the second column of Table 4 documents, since 1994
the standard deviation of core CPI inflation does not
appear to have been lower in IT economies relative
to non-IT economies. Indeed, each IT economy has
had higher inflation variance over this period than
Denmark, the euro area, Japan, and the United States.
Using the autoregression in (2), the volatility
of inflation can be decomposed into two sources:
one due to the variance of the shocks to the auto
regression and one due to the propagation of
shocks through the autoregressive dynamics. The
final column in Table 4 gives one measure of this
decomposition—the ratio of the total variance of
the inflation series to the variance of shocks to the
autoregression. With the exception of Australia,
these ratios are only slightly above unity in the IT
countries, consistent with a white noise process for
the inflation series. By contrast, this ratio is near or
above 2.0 in the euro area, Japan, and the United
States. Thus, it appears that the volatility of inflation
in these non-IT economies contains a substantial
propagation component, while in the IT countries
the initial impact of shocks accounts for nearly all
inflation variance. That overall variance is roughly

FEDERAL R ESERVE BANK OF ST. LOUIS

similar in the two economies suggests that shocks
to inflation in IT countries have been large relative
to non-IT countries, and, had these economies not
experienced low levels of inflation persistence,
inflation volatility would have been even higher.

5. THE CHARACTERISTICS OF
INFLATION TARGETING IN
EMERGING MARKET ECONOMIES
In recent years, a growing number of EMEs have
adopted IT as the main anchor guiding monetary
policy.12 During the mid-to-late 1990s, monetary
aggregates became increasingly difficult to gauge,
due to instability in money demand, while financial
crises contributed to the widespread collapse of
exchange rate pegs. As a result, many EMEs turned
to IT as the only nominal anchor still viable.
The seminal papers on IT in emerging markets
were aimed at identifying the prerequisites for successful adoption, based on the experience of industrial countries.13 Subsequent analysis centered on
special issues for EMEs, including fiscal dominance
and the role of exchange rates.14 Finally, a number
of recent studies have analyzed the initial effects of
IT for EMEs in Eastern Europe and Latin America.15
This section investigates the experience of EMEs,
focusing on the circumstances under which they
adopted IT and on some of the distinctive features
and problems in the emerging market context. The
next section considers the effects of IT in these
economies, focusing in particular on the impact on
inflation expectations.
Chile introduced IT in 1991. After gaining independence in 1990, the central bank of Chile faced a
significant increase in inflation following expansionary policies in 1989 and the oil price spike related
to the first Gulf War. Having already unsuccessfully
experienced two exchange rate–based stabilization
12

Since Chile and Israel first introduced IT in the early 1990s, EMEs
that have formally instituted IT include Brazil, Colombia, the Czech
Republic, Hungary, South Korea, Mexico, Peru, the Philippines, Poland,
South Africa, and Thailand. See Table 5 for details.

13

See, for instance, Masson, Savastano, and Sharma (1997) and Agenor
(2000).

14

See Amato and Gerlach (2002), Blejer et al. (2000), Cukierman, Miller,
and Neyapti (2002), and Mishkin (2000).

15

Fraga, Goldfajn, and Minella (2004) is a comprehensive study of the
performance of IT in EMEs, with special attention to the Brazilian case.
For further lessons from Latin America, see Calderon and SchmidtHebbel (2003), Corbo and Schmidt-Hebbel (2001), and Mishkin and
Savastano (2002). For analysis of IT in transition economies, see Jonas
and Mishkin (2003).

Levin, Natalucci, Piger

programs in the past and with monetary aggregates
difficult to control due to instability in money
demand, IT was the only viable alternative. A key
feature of the Chilean experience has been the gradual approach to disinflation, which has produced
low inflation without suffering excessively large output costs. Chile had an exchange rate band around
a crawling peg until August 1999; it has since adopted
a fully floating exchange rate regime. IT in Chile has
been generally successful in bringing down inflation,
even though a strong fiscal position and a sound
financial system played an important role in supporting this performance.
Israel’s monetary policy framework has been
centered on the coexistence of two nominal goals,
the inflation target and a crawling exchange rate
band, supported by one instrument, the interest rate.
Following the 1985 stabilization program, characterized by a fixed but adjustable nominal exchange
rate, at the beginning of 1992 Israel adopted an
explicit inflation target. Inflation has been successfully reduced from double digits to practically zero.
However, the emergence of a conflict between the
two nominal objectives often required sterilized
foreign exchange intervention, with associated quasifiscal costs and weakening of the central bank’s
credibility. With the widening of the band to 36 percent and the setting of a clear hierarchy of priorities,
this conflict appears now to have lessened.
The successful experience of Chile and Israel
paved the way for the adoption of IT in other EMEs.
In East Asia, the first country to introduce this monetary policy framework was South Korea. Before the
adoption of IT in 1998, monetary policy had been
conducted by deciding on monetary aggregates as
an intermediate target. However, following rapid
structural changes experienced by financial markets
in the 1990s, the M2 aggregate began to show unstable movements. With the 1997 financial crisis forcing
the abandonment of the exchange rate peg, Korea
turned to IT as the only nominal anchor for monetary
policy still available. Thailand and the Philippines
shared a similar experience and adopted IT in 2000
and 2002, respectively.
A trend toward more-flexible exchange rates
has also been observed in some of the transition
economies of Central and Eastern Europe. Following price liberalization and exchange rate devaluation in the early years of transition, most countries
resorted to exchange rate pegs to stabilize their price
levels. However, a sharp appreciation of the real
exchange rate generated large balance-of-payment
J U LY / A U G U S T 2 0 0 4

63

REVIEW

Levin, Natalucci, Piger

problems, forcing some countries to abandon the
peg and float their currencies: the Czech Republic
in May 1997 after currency turbulence and the
Slovak Republic and Poland in 1998. Hungary never
adopted a fully floating exchange rate, but has been
living with a ±15 percent exchange rate band since
2001.
In need of a new nominal anchor, the Czech
Republic was the first country to adopt IT at the
beginning of 1998. Poland followed suit in mid-1998.
In contrast, Hungary’s move to IT has been more
gradual, with a progressive widening of the exchange
rate band and the introduction of IT in 2001.16
Mexico and Brazil were the first (and the largest)
Latin American countries to introduce an IT regime.
In Mexico, after floating the peso in December 1994,
the central bank tried to maintain its monetary targeting regime for a few years. Due to the unreliability
of the relationship between the monetary base and
inflation, however, the stance of monetary policy
was difficult to assess and the Bank of Mexico lacked
a nominal anchor to guide inflation expectations.
IT was the natural candidate: It was introduced
gradually and adopted in 1999.
In Brazil, the real plan introduced in 1994 successfully reduced inflation from above 2000 percent
to 1.5 percent in 1998. However, the Brazilian government was not as successful in implementing muchneeded fiscal reforms. Following concerns about
the fiscal balance, the real came under speculative
attack at the end of 1998 and collapsed in January
1999. The central bank acknowledged the need to
put in place a nominal anchor and, after sharply
raising interest rates to slow the fall of the currency,
introduced an IT regime in June 1999.
In Colombia and Peru, some characteristics of
an IT regime were already present in the first half
of the 1990s.17 However, many important features
were missing, including the publication of inflation
reports, multi-year targets for inflation, transparency,
etc. We therefore set the IT adoption date in
September 1999 for Colombia and January 2002
for Peru.18
16

The ±15 percent exchange rate band was introduced in May 2001,
and the rate of crawl was eliminate only in October 2001.

17

These include some degree of central bank independence, the
announcement of explicit numerical targets for the one-year-ahead
inflation rate (often in conjunction with the government economic
program), etc.

18

The experience of Colombia is similar to Brazil’s, where, after unsuccessfully defending the exchange rate band in September 1999, the
authorities let the currency float and adopted IT as the nominal anchor.

64

J U LY / A U G U S T 2 0 0 4

In South Africa, following financial liberalization
and other structural developments in the 1990s,
the changing relationship between growth in money
supply, output, and prices made explicit monetary
growth targets less and less useful. In 1998, M3
growth guidelines started to be accompanied by
informal targets for inflation, and in early 2000 a
formal IT framework was finally introduced.19
After having investigated the circumstances
under which these EMEs adopted IT, we are now
interested in the main characteristics of these
regimes, particularly compared with industrial countries. Table 5 summarizes the main features of IT in
EMEs. There are several points worth noting. First
of all, the current inflation targets are relatively low
and not much higher than they are in industrial
countries.20 The experience of EMEs with respect
to the disinflation process has varied, with some
countries following a gradual approach and others
being more aggressive. Overall, as shown in Figure 5,
most of the countries have been successful in bringing down inflation from double digits to single digits.21
But what should the appropriate target level be for
EMEs? It is sometimes argued that central banks in
EMEs should aim for somewhat higher rates of inflation than industrial economies, due to the presence
of the Balassa-Samuelson effect. This is still an open
question.
Second, EMEs seem split in choosing either a
target point with a range around it or a target range.
When countries choose a target point, the range is
always ±1 percent.22 Instead, when they choose a
target range, this can be as narrow as 1 percent and
as wide as 3.5 percent.23 Only one country, Thailand,
has chosen a target range with a lower threshold of
0 percent. It remains an open question whether a
For Peru, we refer to the January 2002 Monetary Program, which
states that “As of this year, the Central Reserve Bank of Peru (BCRP)
has adopted an Explicit Inflation Targeting system.” Other authors
(e.g., Fraga, Goldfajn, and Minella, 2004) set the adoption date in 1994.
19

See Casteleijn (2001).

20

With the exception of Brazil, the inflation targets are included in a
range between 0 and 6 percent. If we also exclude Colombia, the
Philippines, and South Africa, the targets are centered on 2 to 3 percent.

21

CPI 12-month percent changes in November 2003 were at or below
5 percent in 10 of 13 countries. The exceptions are Brazil, Colombia,
and Hungary, with inflation at 11.1, 6.1, and 5.6 percent, respectively.

22

Only one country, Brazil, has a range of ±2.5 percent (for 2004 and
2005).

23

Brazil, again, is the only exception to this regularity. See Table 5 for
details.

FEDERAL R ESERVE BANK OF ST. LOUIS

Levin, Natalucci, Piger

Figure 5
Inflation Rates and Targets (quarterly inflation rates are year over year)
Chile

Percent
30

Czech Republic

Percent
14
12

25

10
20

8

15

6
4

10

2
5

1990

1992

1994

1996

1998

2000

2002

0

0
1990

1992

1994

1996

1998

2000

2002

–2

Inflation targets through the end of 2001 are for net inflation.
Israel

Percent
25

South Africa

Percent
20

20

15

15
10
10
5

5

1990

1992

1994

1996

1998

2000

South Korea

2002

0

Percent
12

1990

1992

1994

1996

1998

2000

Thailand

2002

0

Percent
12
10

10

8

8

6
6
4
4

2

2

1990

1992

1994

1996

1998

2000

2002

0

0

1990

1992

1994

1996

1998

2000

2002

–2

Since 2000, core inflation (the dotted line) has been targeted;
previously, CPI (the solid line) was targeted.

J U LY / A U G U S T 2 0 0 4

65

REVIEW

Levin, Natalucci, Piger

Figure 5 cont’d
Inflation Rates and Targets (quarterly inflation rates are year over year)
Percent
30

Brazil

Percent
35

Colombia

30

25

25

20

20
15
15
10

10

5

1990

1992

1994

1996

1998

2000

2002

0

5

1990

1992

1994

1996

1998

2000

2002

0

The band of ± 2 percent was made explicit in 2003.
Hungary

Percent
40

Mexico

Percent
50

40

30

30
20
20
10

1990

1992

1994

1996

1998

2000

2002

0

10

1990

1992

1994

1996

1998

2000

The lower band was made explicit in 2003.
Percent
50

Poland

40

30

20

10

1990

66

1992

J U LY / A U G U S T 2 0 0 4

1994

1996

1998

2000

2002

0

2002

0

FEDERAL R ESERVE BANK OF ST. LOUIS

Levin, Natalucci, Piger

Figure 5 cont’d
Inflation Rates and Targets (quarterly inflation rates are year over year)
8

10
8

6
6
4

4
2

2
0

1

0

1

2

1

4

1

6

1

8

2000

2002

–2

1

0

1

2

1

4

1

6

1

8

2000

2002

8

0

12
10

6
8
4

6
4

2

2
0
0

1

0

1

2

1

4

1

6

1

8

2000

2002

–2

1

0

1

2

1

4

1

6

1

8

2000

2002

–2

12

8

4

1

0

1

2

1

4

1

6

1

1

8

2000

2002
2

0
W

± 1
q
1

J U LY / A U G U S T 2 0 0 4

67

REVIEW

Levin, Natalucci, Piger

Table 5
Features of IT Regimes in Developing Countries
Brazil

Chile

Colombia

Czech Republic

Hungary

Israel

Jun 1999

Jan 1991

Sep 1999

Jan 1998

Aug 2001

Jan 1992

Current target

1.5-8.5

2-4 centered at 3

6

3-5 declining to 2-4

3.5 ± 1

1-3

Target duration

5.5 ± 2.5 (2004)
3.5 ± 2.5 (2005)

Medium term

5-6 (2004)

Through
Dec 2005

3.5 ±1 (2004)
2 (long term)

2003 onward

CPI

CPI

CPI

CPI

Date first issued

Inflation
measure

National consumer
CPI; central
price index
bank monitors
(IPCA): a measure
core inflation
of inflation in
(which excludes
9 metro areas
vegetable, fruit,
plus 2 other
and fuel prices
urban areas

Target
announcement

Set by National
Monetary Council,
composed by
finance minister,
planning minister,
and central
bank president

Central bank in
consultation
with
government

Jointly by
government
and central
bank

Central bank

Central bank

Minister of finance
in consultation
with prime
minister and
governor
of central bank

Inflation report

Yes

Yes

Yes

Yes

Yes

Yes

Published forecast

Yes

Yes

Yes

Yes

Yes

No

Other objectives

—

—

—

—

±15% band
around parity
with Euro

±36% crawling
band around
parity with a
currency basket
representing
Israel’s foreign trade

Price stability,
sound financial
system

Price stability,
functioning
payments system

Price stability

Price stability

Price stability

Price stability

Letter from
central bank
president to
minister of finance
if target breached

—

—

—

—

Public explanation
when deviations
from target
are greater than
±1%

Mandate

Other features

68

J U LY / A U G U S T 2 0 0 4

FEDERAL R ESERVE BANK OF ST. LOUIS

Levin, Natalucci, Piger

Korea

Mexico

Peru

Philippines

Poland

South Africa

Thailand

Apr 1998

Jan 1999

Jan 2002

Jan 2002

Jun 1998

Feb 2000

May 2000

3 ±1

3 ±1

2.5 ± 1

4.5-5.5

3 ±1

3-6

0-3.5

2.5-3.5 (average
2004-2006)

Around 3
(medium term)

2004

4-5 (2004)

2.5 ±1
(medium term)

2004

2004

Core inflation
(CPI inflation
minus non-cereal
agricultural
products and
petroleumbased products)

CPI

CPI

CPI, although
four core inflation
measures are
monitored by the
central bank

CPI

CPI (excluding
mortgage
interest costs)

Core CPI
(excluding raw
food and energy
prices)

Central bank in
consultation with
government

Central bank

Central bank

Set and
announced jointly
by central bank
and government

Central bank

Central bank

Government in
consultation
with central
bank

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

No

No

Yes

—

—

Foreign exchange
operations

—

—

—

—

Price stability,
sound financial
system

Price stability

Price stability

—

Price stability,
Price stability
sound financial
system, functioning
payments system

—

—

Price stability
Price stability,
conducive to
necessary in
balanced and
building the
suitable
permanent
economic growth, foundation of
monetary stability,
long-term
convertibility
economic
of currency
growth
Letter from
central bank
governor to
president when
target breached

Public
explanation when
target breached

J U LY / A U G U S T 2 0 0 4

69

REVIEW

Levin, Natalucci, Piger

target point or a target range should be chosen. In
favor of the target point, it should be noted that the
point appears to be more effective in focalizing
inflation expectations. And the range around it still
allows for some flexibility in the event of forecast
errors or unexpected events. In the presence of a
target range, instead, the thresholds sometimes
seem to be assuming life on their own.
Third, following the earlier experience of industrial countries, most EMEs moved away from oneyear-ahead inflation targets and adopted multi-year
targets or some definition of a medium-term target.24
This can be interpreted as a sign that the disinflation
process from high levels of inflation has come close
to an end, forcing these countries to “think mediumterm” and develop a more operational concept of
price stability.
Fourth, most EMEs target the CPI because it is
well understood by the public and quickly available.25
Despite this, emerging and advanced countries have
at least two main differences in their respective CPI
baskets. First, the share of food is larger in EMEs.
This implies a more volatile CPI, since food prices
are related to weather conditions and therefore tend
to move more unpredictably. Second, regulated
prices have a greater impact in EMEs, especially
during the early years of the disinflation process.
Consequently, it is more difficult for the central bank
to effectively control inflation, with potential damage to the central bank’s credibility.26 However,
while targeting core inflation would probably be
more appropriate, a measure of inflation that disregards food and regulated prices might not reflect
the cost of living, putting the public support for an
independent central bank at risk.
Finally, EMEs seem to be moving away from
previous attempts to control two objectives, inflation and the exchange rate, with one instrument.27
24

This is true for Brazil, Chile, the Czech Republic, Hungary, Israel,
South Korea, Mexico, and Poland.

25

Exceptions are Brazil, South Korea, South Africa, and Thailand. Other
countries (Chile and the Philippines) monitor some measures of core
inflation.

26

27

Two broader issues are related to the central banks’ ability to control
inflation in EMEs. One has to do with the Balassa-Samuelson effect,
which implies an appreciation of the real exchange rate either via
higher inflation or via an appreciation of the nominal exchange rate.
The second issue has to do with the difficulty of forecasting inflation.
This is true after a regime change, during disinflation from high inflation levels, and because of EMEs’ sensitivity to commodity prices and
disproportionate dependence on capital flows.
A strategy of dual objectives was originally adopted in some EMEs to
speed up the disinflation process. The introduction of exchange rate

70

J U LY / A U G U S T 2 0 0 4

In fact, only Hungary and Israel still have a band
for the nominal exchange rate. There are several
reasons why EMEs may want to pay greater attention
to exchange rates than industrial countries. First,
with large shocks and sizable capital flows, neglecting the exchange rate may generate unwelcome
volatility. Second, in countries with historically high
inflation, the exchange rate may work as a focal
point for inflation expectations.28 Third, since firms
and governments in EMEs borrow mainly in foreign
currency, large depreciations may increase the
burden of foreign-denominated debt, producing a
massive deterioration of balance sheets and increasing the risks of a financial crisis.29 However, most
EMEs have decided to focus their efforts primarily
on controlling inflation and have abandoned the
idea of managing extensively the exchange rate,
which can be interpreted as an additional sign of
their intention to embrace a fully fledged IT regime.

6. THE EFFECTS OF INFLATION
TARGETING IN EMERGING MARKET
ECONOMIES
In considering the effects of IT in EMEs, we begin
by focusing on inflation expectations. For each
country for which data are available, Figure 6 plots
(i) realized inflation (measured as Q4/Q4); (ii) oneyear-ahead expected inflation (on a Q4/Q4 basis),
where the expectation is formed in the fourth quarter
of the current year; and (iii) long-run (6 to 10 years)
inflation expectations, where the expectation is
formed in the fourth quarter of the current year.
Inflation expectations are again measured based
on surveys conducted by Consensus Economics.
The figure contains data for three years before and
after the adoption date. The data used in creating
Figure 6 are shown in Appendix Table A3.
We begin by considering long-term inflation
expectations. The main result is that, as in industrial
countries, IT does not seem to have had a large initial
impact on long-term expected inflation. In other
flexibility became necessary to resolve the tension between maintaining
the disinflationary momentum and guarding against a loss of competitiveness. As the disinflation process continued, the bands were typically
broadened and subsequently abandoned as they became a source of
policy conflict, undermining the credibility of the inflation target. The
experience of Hungary in January, June, and December 2003 highlights
the risks of combining IT and exchange rate management in periods
of speculative attacks and large swings in market sentiment.
28

Depreciations have historically tended to have larger inflationary
effects in EMEs, as pass-through effects have been faster.

29

For the discussion on the composition of the CPI basket and the role
of the exchange rate in EMEs, we relied on Amato and Gerlach (2002).

FEDERAL R ESERVE BANK OF ST. LOUIS

Levin, Natalucci, Piger

Figure 6
Event Study: IT in EMEs
Korea
12
IT Adoption
Inflation Rate
Inflation 1 Year Ahead
Inflation 6-10 Years Ahead
Ex Ante Real Interest Rate
9

6

3

0
1995

1996

1997

1998

1999

2000

2001

Brazil
40
IT Adoption

30

20

10

0
1996

1997

1998

1999

2000

2001

2002

J U LY / A U G U S T 2 0 0 4

71

REVIEW

Levin, Natalucci, Piger

Figure 6 cont’d
Event Study: IT in EMEs
Mexico
30.0
IT Adoption
Inflation Rate
Inflation 1 Year Ahead
Inflation 6-10 Years Ahead
Ex Ante Real Interest Rate
22.5

15.0

7.5

0.0
1996

1997

1998

1999

2000

2001

2002

Hungary
12
IT Adoption

9

6

3

0
1998

72

J U LY / A U G U S T 2 0 0 4

1999

2000

2001

2002

2003

2004

FEDERAL R ESERVE BANK OF ST. LOUIS

Levin, Natalucci, Piger

Figure 6 cont’d
Event Study: IT in EMEs
Thailand
12
IT Adoption
Inflation Rate
Inflation 1 Year Ahead
Inflation 6-10 Years Ahead
Ex Ante Real Interest Rate
8

4

0

–4
1997

1998

1999

words, long-term inflation expectations did not
change dramatically at the time of the adoption of
IT. Consider Brazil: Inflation expectations were 2.4
percent at the end of 1998, when the real came under
attack, down from nearly 6 percent in 1996. They
were up to only 3.4 in 1999, when IT was introduced
after the collapse of the currency, even though actual
inflation jumped to 8.2 percent from 1.7 percent in
the previous year. Inflation expectations continued
to decline in 2000, down to 3.1 percent, with actual
inflation still above that level, at 6.1 percent. Inflation
expectations rose only slightly afterwards, up to
4.4 percent, well below actual inflation. The main
point, therefore, is that inflation expectations in
Brazil started to decline before the adoption of IT
and continued to do so afterwards, edging up again
2 years later, but always remaining below actual
inflation.
A similar path can be observed in other countries. In South Korea, inflation expectations have
been declining since 1995, well before the adoption
of IT, and continued to fall smoothly, at small decrements, through 2001, down 1 percent in total. Actual
inflation rose only 1 percent after the financial crisis,
in 1998, but dropped to 1.3 percent in 1999, well
below long-term inflation expectations. In 2001,
actual inflation was at a level consistent with long-

2000

2001

2002

2003

term expectations. In Mexico, apart from 1998, inflation expectations dropped dramatically, from 10.4
percent in 1996 to 7.5 percent in 1999 and 3.4 percent in 2002. The introduction of IT does not seem
to have affected significantly this downward trend
in inflation expectations. Moreover, inflation expectations have been consistently below actual inflation,
even immediately after the 1994-95 crisis, when
the difference was almost 20 percent. At the end of
2002, long-term expectations were 2 percent below
actual inflation. In Thailand, inflation expectations
have declined since 1998, with a noticeable drop
in 2000, when IT was adopted, but have been well
above actual inflation since 1999. Finally, inflation
expectations in Hungary were coming down before
IT was introduced and actually rose the year of the
adoption.30 However, they remained stable, at
around 3 percent in 2003, even though actual inflation rose almost 1 percentage point.
30

For Colombia, we don’t have inflation expectations for the year after
the adoption of IT, but there is still a clear downward trend beginning
in 1997. In Peru, the decline in long-term inflation expectations began
before IT was introduced in 2002, but was very gradual. For the Czech
Republic and Poland we don’t have data available for the years before
adoption of the IT. However, inflation expectations have been declining
since the adoption of IT. Finally, we don’t have any data available for
Chile and Israel (the early adopters) or for the Philippines and South
Africa.

J U LY / A U G U S T 2 0 0 4

73

REVIEW

Levin, Natalucci, Piger

Table 6
Relative Success in Hitting Inflation Targets
(standard deviation from midpoint)
Country

Standard deviation

Brazil

5.4

Chile

1.9

Columbia

2.1

Czech Republic

3.1

Hungary

1.1

Israel

2.6

Mexico

2.1

Poland

2.7

South Africa

3.6

Korea

1.3

Thailand

0.7

Australia

1.7

Canada

1.2

New Zealand

1.6

Sweden

1.3

United Kingdom

0.9

NOTE: Inflation is measured as a quarterly, annualized rate.
For Columbia, inflation deviations are based on CPI inflation,
although the target is based on net inflation through 2001. In
accordance with the target, inflation deviations for South Korea
are based on CPI inflation through 1999 and on core inflation
thereafter.

What about short-term inflation expectations?
Is there any evidence that the introduction of IT
lowered one-year-ahead expectations? The conclusion is similar to the case of long-term inflation
expectations: There is no evidence of any dramatic
reduction in short-term inflation expectations, neither for the year IT was introduced nor for the following year. There seems to be, instead, a gradual decline
of these expectations over time, with differences
on a country-by-country basis.31
31

For example, in Brazil, in line with long-term inflation expectations,
short-term expectations rose the year of the introduction of IT, declined
the following year, and rose again in the next couple of years. In Hungary,
short-term inflation expectations declined both the year IT was introduced and the following year. However, a downward trend was already
evident in the previous three years. A similar story holds for Mexico,
with short-term expectations gradually declining over time, well before
the introduction of IT. In Korea and Thailand, short-term expectations
actually rose the year IT was adopted and dropped significantly the year
after. Interestingly, this was the only year of such an increase, with
both the previous and the following three years showing declines.

74

J U LY / A U G U S T 2 0 0 4

In summary, the evidence from inflation expectations suggests that, while expectations declined
when IT was introduced and continued to do so
subsequently, the downward trend was evident even
before the switch to IT, in line with the experience
of industrial countries. This does not necessarily
mean that IT was ineffective, as it is plausible that,
in the absence of IT, “bad” monetary policies could
have offset previous gains in reducing inflation.
An alternative way to evaluate the medium-term
performance of IT in EMEs would be to calculate
sacrifice ratios for these countries, along the line of
similar studies for industrial countries. However,
EMEs are characterized by rapid structural changes,
making the estimate of potential output extremely
difficult and maybe even unreliable. One possibility
is to look at short-term ex ante real rates (shown in
Figure 6 and Appendix Table A3). Consider Brazil,
for example. Short-term real rates were very high
before the introduction of IT, at almost 36 percent
in 1998, came down to 13.4 percent in 1999, but
remained around that level for the following three
years. Monetary policy was very tight, and this makes
it more difficult to evaluate the performance of IT
as a monetary policy framework. In Mexico, shortterm real rates were very high in 1998, but declined
substantially the year IT was introduced, down to
nearly 5 percent. After rising in 2000, they were
around 1 percent in 2001 and 2002. In this case, it
seems reasonable to conclude that the successful
reduction of inflation cannot be entirely attributed
to tight monetary policy, leaving some scope for
crediting IT. This is even more evident in Korea,
where short-term real rates dropped the year of
the introduction of IT, from nearly 12 percent to
2.6 percent, and remained low afterward, and in
Thailand, where real rates were negative even the
two years before the introduction of IT. In summary,
while in some countries real rates were very high
when IT was introduced, in other countries real
rates were low and inflation was still successfully
reduced.
In EMEs the adoption of IT has been frequently
associated with overshooting and undershooting
of the targets. An alternative way to evaluate the
medium-term performance of the IT framework in
EMEs is to look at the frequency of overshooting and
undershooting. Table 6 shows the standard deviation
of inflation from the midpoint of the target range
for each of the countries considered in Figure 5.
Not surprisingly, industrial countries generally display a lower standard deviation than EMEs. Among

FEDERAL R ESERVE BANK OF ST. LOUIS

EMEs, Brazil is the worst performer, followed by
South Africa, while South Korea and Thailand are the
best performers, with standard deviations even lower
than that of Australia. Possible explanations for
the higher standard deviation of inflation in EMEs
include the difficulty of controlling and forecasting
inflation in the developing world, the larger shocks
EMEs face, and the lower credibility central banks
have in countries with a history of high inflation.
In conclusion, the record to date suggests that
inflation targeters in emerging markets have been
relatively successful in reducing inflation, although
the record is still fairly short for most of the countries. It is still not completely obvious, however, the
extent to which this reduction can be credited
entirely to IT as a monetary policy framework. It
might be the case that part of the success of IT in
EMEs is attributable to the global downward trend
in inflation rates. It remains also to be seen whether
the fairly strong performance of these countries will
be sustained over a longer horizon.

7. CONCLUSION
Our analysis of the past decade of experience
for the industrial countries suggests that IT has
played a role in anchoring inflation expectations
and in reducing inflation persistence. Of course,
because we have focused on reduced-form evidence,
we have not addressed the extent to which certain
country-specific factors may account for the differences we have documented across IT and non-IT
economies. For example, many of the IT countries
in our sample are small, open economies, which
might be expected to have very different inflation
dynamics from the large, mostly closed economies
that dominate our non-IT sample.
Nevertheless, our results are broadly consistent
with the implications of the expectations-augmented
Phillips curve:
(3)

π t = πˆ t +1 + φ yt + ε t ,

where π̂ t+1 is the one-period-ahead forecast of inflation, yt is the current output gap, and εt is an aggregate supply shock. When the central bank has an
transparent and credible inflation target, π*, then
the private sector’s inflation forecast corresponds
to π̂ t+j=π* at some reasonable forecast horizon, j.
In this case, actual inflation will depend on expected
output gaps over the next j periods and on the current aggregate supply shock. Thus, inflation will tend
to exhibit relatively little intrinsic persistence in

Levin, Natalucci, Piger

response to transitory supply shocks; the observed
degree of inflation persistence may depend largely
on the persistence of output gap fluctuations. As a
result, under IT, a key challenge for the central bank
may be to keep output close to potential by moving
promptly to offset aggregate demand shocks.
In contrast, if the central bank’s inflation objective is not transparent or credible, the private sector’s
rational forecast of medium-to-long-run inflation
will depend on the recent behavior of actual inflation
(cf. Erceg and Levin, 2003). For the simplest case in
which π̂t+1=π̂t–1, it is evident that inflation will tend
to exhibit a high degree of intrinsic persistence,
even in response to temporary supply shocks or
fluctuations in aggregate demand.32
Our investigation of the early experience with
IT in EMEs confirms that—as in the industrial countries—the adoption of IT has generally not been
associated with an instantaneous adjustment of
inflation expectations. Furthermore, while most of
these EMEs have succeeded in reducing average inflation to very low levels, the volatility of inflation has
remained quite high, with relatively frequent overshooting and undershooting of the target bands.
Such volatility is not necessarily surprising, given
that most of the EMEs are small and highly sensitive
to global economic fluctuations. Thus, additional
research and experience will be helpful in finetuning the implementation of IT and ensuring its
positive contribution to macroeconomic stability.

REFERENCES
Agenor, Pierre-Richard. “Monetary Policy under Flexible
Exchange Rates: An Introduction to Inflation Targeting.”
Unpublished manuscript, 2000.
Akaike, H. “Information Theory and an Extension of the
Maximum Likelihood Principal,” in B.N. Petrov and F. Csaki,
eds., Second International Symposium on Information
Theory. Budapest: Akademia Kiado, 1973, pp. 267-81.
Almeida, Alavro and Goodhart, Charles A.E. “Does the
Adoption of Inflation Targets Affect Central Bank
Behaviour?” Unpublished manuscript, 1998.
Amato, Jeffrey D. and Gerlach, Stefan. “Inflation Targeting
in Emerging Market and Transition Economies: Lessons
after a Decade.” European Economic Review, April 2002,
46(4-5), pp. 781-90.
32

See also Orphanides and Williams (forthcoming).

J U LY / A U G U S T 2 0 0 4

75

Levin, Natalucci, Piger

Ammer, John and Freeman, Richard T. “Inflation Targeting
in the 1990s: The Experiences of New Zealand, Canada,
and the United Kingdom.” Journal of Economics and
Business, May 1995, 47(2), pp. 165-92.
Andrews, Donald W.K. and Chen, Hong-Yuan. “Approximately
Median-Unbiased Estimation of Autoregressive Models.”
Journal of Business and Economic Statistics, April 1994,
12(1), pp. 187-204.
Ball, Laurence and Sheridan, Niahm. “Does Inflation
Targeting Matter?” in Ben S. Bernanke and Michael
Woodford, eds., The Inflation Targeting Debate. Chicago:
University of Chicago Press (forthcoming).
Barsky, Robert B. “The Fisher Hypothesis and the
Forecastability and Persistence of Inflation.” Journal of
Monetary Economics, January 1987, 19(1), pp. 3-24.
Batini, Nicoletta. “Euro Area Inflation Persistence.” Working
Paper No. 201, European Central Bank, 2002.
Benati, Luca. “Investigating Inflation Persistence Across
Monetary Regimes.” Unpublished manuscript, Bank of
England, 2002.
Benati, Luca. “Evolving Post-World War II U.K. Economic
Performance.” Unpublished manuscript, Bank of England,
2003.
Bernanke, Ben S.; Laubach, Thomas; Mishkin, Frederic and
Posen, Adam S. Inflation Targeting: Lessons from the
International Experience. Princeton, NJ: Princeton
University Press, 1999.

REVIEW
Finance and Economics Discussion Series Paper No.
2003-15, Board of Governors of the Federal Reserve
System, 2003.
Brainard, William C. and Perry, George. “Making Policy in a
Changing World,” in G. Perry and J. Tobin, eds., Economic
Events, Ideas, and Policies: The 1960s and After.
Washington, DC: Brookings Institution Press, 2000.
Calderon, César A. and Schmidt-Hebbel, Klaus. “Macroeconomic Policies and Performance in Latin America.”
Working Paper No. 217, Central Bank of Chile, 2003.
Casteleijn, Anton J.H. “South Africa’s Monetary Policy
Framework.” Working paper, Reserve Bank of South
Africa, 2001.
Castelnuovo, Efrem; Nicoletti-Altimari, Sergio and
Palenzuela, Diego. “Definition of Price Stability, Range
and Point Inflation Targets: The Anchoring of Long-Term
Inflation Expectations,” in O. Issing, ed., Background
Studies for the ECB’s Evaluation of Its Monetary Policy
Strategy. Frankfurt-am-Main, Germany: European Central
Bank, 2003.
Cecchetti, Stephen G. and Ehrmann, Michael. “Does Inflation
Targeting Increase Output Volatility? An International
Comparison of Policymakers’ Preferences and Outcomes.”
NBER Working Paper 7426, National Bureau of Economic
Research, 1999.
Cogley, Timothy and Sargent, Thomas J. “Evolving PostWorld War II U.S. Inflation Dynamics,” in B.S. Bernanke
and K. Rogoff, eds., NBER Macroeconomics Annual 2001.
16(1). Cambridge, MA: MIT Press, 2002, pp. 331-72.

Bernanke, Ben S. and Kuttner, Kenneth N. “What Explains
the Stock Market’s Reaction to Federal Reserve Policy?”
Staff Report No. 174, Federal Reserve Bank of New York,
2003.

Cogley, Timothy and Sargent, Thomas J. “Drifts and
Volatilities: Monetary Policies and Outcomes in the
Postwar U.S.” Working paper, 2003.

Bernanke, Ben S. and Mishkin, Frederic. “Inflation Targeting:
A New Framework for Monetary Policy?” Journal of
Economic Perspectives, Spring 1997, 11(2), pp. 97-116.

Corbo, Vittorio and Schmidt-Hebbel, Klaus. “Inflation
Targeting in Latin America.” Working Paper No. 105,
Central Bank of Chile, 2001.

Blejer, Mario I.; Ize, Alain; Leone, Alfredo M. and Werlang,
Sergio. Inflation Targeting in Practice: Strategic and
Operational Issues and Application to Emerging Market
Economies. Washington, DC: International Monetary
Fund, 2000; www.imf.org/external/pubs/ft/seminar/2000/
targets/strach1.pdf.

Corbo, Vittorio; Landerretche, Oscar and Schmidt-Hebbel,
Klaus. “Assessing Inflation Targeting after a Decade of
World Experience.” Working Paper No. 51, Austrian
National Bank, 2001.

Bonfim, Antulio N. “Monetary Policy and the Yield Curve.”
76

J U LY / A U G U S T 2 0 0 4

Cukierman, Alex; Miller, Geoffrey P. and Neyapti, Bilin.
“Central Bank Reform, Liberalization and Inflation in
Transition Economies—An International Perspective.”

FEDERAL R ESERVE BANK OF ST. LOUIS

Journal of Monetary Economics, May 2002, 49(2), pp.
237-64.
Dueker, Michael J. and Fischer, Andreas. “Do Inflation
Targets Redefine Central Bank Inflation Preferences?
Results from an Indicator Model,” in K. Alders, K. Koedijk
and C. Winder, eds., Monetary Policy in a Converging
Europe. Amsterdam: Kluwer Academic Publishers, 1996.
Erceg, Christopher J. and Levin, Andrew T. “Imperfect
Credibility and Inflation Persistence.” Journal of Monetary
Economics, May 2003, 50(4), pp. 915-44.
Evans, Martin and Wachtel, Paul. “Inflation Regimes and
the Sources of Inflation Uncertainty. “ Journal of Money,
Credit, and Banking, August 1993, 25(3), pp. 475-511.
Fraga, Arminio; Goldfajn, Ilan and Minella, Andre. “Inflation
Targeting in Emerging Market Economies,” in M. Gertler
and K. Rogoff, eds., NBER Macroeconomics Annual 2003.
Volume 18. Cambridge, MA: MIT Press (forthcoming).
Franses, Philip H. and Haldrup, Hans. “The Effects of Additive
Outliers on Tests for Unit Roots and Cointegration.”
Journal of Business and Economic Statistics, October 1994,
12(4), pp. 471-78.
Fuhrer, Jeffrey and Moore, George R. “Inflation Persistence.”
Quarterly Journal of Economics, February 1995, 110(1),
pp. 127-59.
Gavin, William T. “Inflation Targeting: Why It Works and
How to Make It Work Better?” Business Economics, April
2004, 39(2), pp. 30-37.
Gertler, Mark. Comment on L. Ball and N. Sheridan, “Does
Inflation Targeting Matter?” in Ben S. Bernanke and
Michael Woodford, eds., The Inflation Targeting Debate.
Chicago: University of Chicago Press (forthcoming).
Goodfriend, Marvin. “Inflation Targeting in the United
States?” in Ben S. Bernanke and Michael Woodford, eds.,
The Inflation Targeting Debate. Chicago: University of
Chicago Press (forthcoming).
Gurkaynak, Refet S.; Sack, Brian and Swanson, Eric. “The
Excess Sensitivity of Long-Term Interest Rates: Evidence
and Implications for Macroeconomic Models.” Finance
and Economic Discussion Series Paper No. 2003-50,
Board of Governors of the Federal Reserve System, 2003.
Hansen, Bruce E. “The Grid Bootstrap and the Autoregressive

Levin, Natalucci, Piger

Model.” The Review of Economics and Statistics, 1999,
81(4), pp. 594-607.
Johnson, David R. “The Effect of Inflation Targeting on the
Behavior of Expected Inflation: Evidence from an 11
Country Panel.” Journal of Monetary Economics, November
2002, 49(8), pp.1493-1519.
Johnson, David R. “The Effect of Inflation Targets on the
Level of Expected Inflation in Five Countries.” Review of
Economics and Statistics, November 2003, 55(4), pp.
1076-81.
Jonas, Jiri and Mishkin, Frederic S. “Inflation Targeting in
Transition Economies: Experience and Prospects.” NBER
Working Paper No. 9667, National Bureau of Economic
Research, 2003.
Kim, Chang-Jin; Nelson, Charles R. and Piger, Jeremy. “The
Less-Volatile U.S. Economy: A Bayesian Investigation of
Timing, Breadth, and Potential Explanations.” Journal of
Business and Economic Statistics, January 2003, 22(10),
pp. 80-93.
Kohn, Donald.L. Comment on M. Goodfriend, “Inflation
Targeting in the United States?” in Ben S. Bernanke and
Michael Woodford, eds., The Inflation Targeting Debate.
Chicago: University of Chicago Press (forthcoming).
Kozicki, Sharon and Tinsley, P.A. “Term Structure Views of
Monetary Policy under Alternative Models of Agent
Expectations.” Journal of Economic Dynamics and Control,
January 2001a, 25(1/2), pp. 149-84.
Kozicki, Sharon and Tinsley, P.A. “Shifting Endpoints in the
Term Structure of Interest Rates.” Journal of Monetary
Economics, June 2001b, 47(3), pp. 613-52.
Kuttner, Kenneth N. and Posen, Adam S. “Does Talk Matter
After All? Inflation Targeting and Central Bank Behavior.”
Federal Reserve Bank of New York, Staff Report No. 88,
1999.
Kuttner, Kenneth N. and Posen, Adam S. “Beyond Bipolar:
A Three-Dimensional Assessment of Monetary Frameworks.” International Journal of Finance and Economics,
October 2001, 6(4), pp. 369-87.
Laubach, Thomas and Posen, Adam S. “Some Comparative
Evidence on the Effectiveness of Inflation Targeting.”
Research Paper No. 9714, Federal Reserve Bank of New
York, 1997.

J U LY / A U G U S T 2 0 0 4

77

Levin, Natalucci, Piger

Leiderman, Leonardo and Svensson, Lars E.O., eds. Inflation
Targets. London: Centre for Economic Policy Research,
1995.
Levin, Andrew T. and Piger, Jeremy. “Is Inflation Persistence
Intrinsic In Industrial Economies?” Working Paper No.
2002-023, Federal Reserve Bank of St. Louis, 2002.
Masson, Paul R.; Savastano, Miguel A. and Sharma, Sunhil.
“The Scope for Inflation Targeting in Developing
Countries.” IMF Working Paper No. 130, International
Monetary Fund, 1997.
Mishkin, Frederic S. “Inflation Targeting in EmergingMarket Countries.” American Economic Review, May 2000,
90(2), pp. 105-09.
Mishkin, Frederic S. and Savastano, Miguel A. “Monetary
Policy Strategies for Emerging Market Countries: Lessons
from Latin America.” Unpublished manuscript, Columbia
University, 2002.
Mishkin, Frederic S. and Schmidt-Hebbel, Klaus. “One
Decade of Inflation Targeting in the World: What Do We
Know and What Do We Need to Know?” in N. Loayza
and R. Soto, eds., A Decade of Inflation Targeting in the
World. Santiago: Central Bank of Chile, 2001, pp. 117-219.
Neumann, Manfred J.M. and von Hagen, Jurgen. “Does
Inflation Targeting Matter?” Federal Reserve Bank of St.
Louis Review, July/August 2002, 84(4), pp. 127-48.
Orphanides, Athanasios and Williams, John C. “Imperfect
Knowledge, Inflation Expectations, and Monetary Policy,”
in Ben S. Bernanke and Michael Woodford, eds., The
Inflation Targeting Debate. Chicago: University of Chicago
Press (forthcoming).
Pivetta, Frederic and Reis, Ricardo A.M.R. “The Persistence
of Inflation in the United States.” Unpublished manuscript,
Harvard University, 2001.
Ravenna, Frederic. “The Impact of Inflation Targeting in
Canada: A Structural Analysis.” Job Market Paper, New
York University, 2000.

78

J U LY / A U G U S T 2 0 0 4

REVIEW
Schaechter, Andrea; Stone, Mark R. and Zelmer, Mark D.
“Adopting Inflation Targeting: Practical Issues for Emerging
Market Countries.” IMF Occasional Paper No. 202,
International Monetary Fund, 2000.
Siklos, Pierre L. “Inflation-Target Design: Changing
Inflation Performance and Persistence in Industrial
Countries.” Federal Reserve Bank of St. Louis Review,
March/April 1999, 81(2), pp. 47-57.
Stock, James H. “Confidence Intervals for the Largest
Autoregressive Root in U.S. Macroeconomic Time Series.”
Journal of Monetary Economics, December 1991, 28(3),
pp. 435-59.
Stock, James H. Comment on T. Cogley and T.J. Sargent,
“Evolving Post-World War II U.S. Inflation Dynamics,” in
B.S. Bernanke and K. Rogoff, eds., NBER Macroeconomics
Annual 2001. 16(1). Cambridge, MA: MIT Press, 2002, pp.
379-87.
Stock, James and Watson, Mark W. “Has the Business Cycle
Changed? Evidence and Explanations.” Federal Reserve
Bank of Kansas City Economic Review (forthcoming).
Svensson, Lars E.O. and Woodford, Michael. “Implementing
Optimal Policy through Inflation-Forecast Targeting,” in
Ben S. Bernanke and Michael Woodford, eds., The Inflation
Targeting Debate. Chicago: University of Chicago Press
(forthcoming).
Taylor, John B. “Low Inflation, Pass-Through, and the
Pricing Power of Firms.” European Economic Review,
2000, 44(7), pp. 1389-1408.
van Dijk, Dick; Osborn, Denise R. and Sensier, Marianne.
“Changes in the Variability of the Business Cycle in the
G7 Countries.” Centre for Growth and Business Cycle
Research Discussion Paper No. 0316, School of
Economics, University of Manchester, 2002.

FEDERAL R ESERVE BANK OF ST. LOUIS

Levin, Natalucci, Piger

Appendix
Table A1
AIC Lag Selection
Country
IT countries
Australia
Canada
New Zealand
Sweden
United Kingdom
Non-IT countries
Denmark
Euro area
France
Germany
Italy
Netherlands
Japan
United States

Lag choice—
core CPI

Lag choice—
total CPI

1
1
1
1
1

1
1
1
1
1

2
2
1
3
1
1
2
4

1
4
3
3
1
1
3
4

Table A2
Alternative Persistence Estimates for Inflation
Core CPI
Country
IT countries
Australia
Canada
New Zealand
Sweden
United Kingdom
Non-IT countries
Denmark
Euro area
France
Germany
Italy
Netherlands
Japan
United States

Median unbiased

Total CPI

Upper 95th percentile

Median unbiased

Upper 95th percentile

0.77

1.05

0.59

0.45
0.43
0.44
0.5

0.73
0.72
0.7
0.77

0.12
0.44
0.28
0.34

0.85
0.46
0.73
0.58
0.64

0.48
0.88
0.79
0.74
0.91
0.53
0.81
1.03

1.07
1.08
1.06
1.09
1.07
0.79
1.10
1.16

–0.05
0.76
0.76
0.65
0.89
0.6
0.5
0.36

0.28
1.24
1.24
1.17
1.07
0.89
1.14
0.87

NOTE: For each country in the sample, this table records the median unbiased estimate and the upper bound of the two-sided 90
percent confidence interval for the sum of the autoregressive coefficients of core and total CPI inflation, estimated over 1994:Q1–2003:Q3.
Estimates were computed based on Hansen (1999), using equation (2).

J U LY / A U G U S T 2 0 0 4

79

REVIEW

Levin, Natalucci, Piger

Table A3
Event Study: IT in EMEs
Years from IT Adoption
–3

–2

–1

Brazil

πt
π̂t(s)
π̂t(l)

0

1

2

3

1999
10.8

5.2

1.7

8.2

6.1

7.4

10.6

8.9

4.6

1.3

5.6

4.8

5.4

7.9

5.7

4.0

2.4

3.4

3.1

3.7

4.4

15.3

31.2

35.9

13.4

11.6

13.7

13.4

πt

11.0

10.6

10.2

7.1

4.8

5.6

NA

π̂t(s)
π̂t(l)

11.4

8.6

7.5

6.2

5.0

5.5

NA

5.8

3.2

2.7

3.7

2.9

3.0

NA

6.3

6.4

3.5

4.1

4.0

7.0

NA

πt

4.4

4.9

4.9

5.9

1.3

2.9

3.0

π̂t(s)
π̂t(l)

5.0

4.6

4.6

4.9

3.3

3.2

2.7

4.1

3.7

3.9

3.6

3.4

3.2

3.1

8.8

11.7

2.6

–0.1

2.0

0.8

rˆt
Hungary

rˆt

2001

Korea

rˆt

1998

NA

Mexico

1999

πt

28.0

17.1

17.2

13.5

8.7

5.1

5.3

π̂t(s)

17.5

12.7

15.4

11.5

7.8

5.5

4.1

π̂t(l)

10.4

7.3

9.4

7.5

5.5

3.7

3.4

9.2

4.7

13.1

4.8

6.3

1.3

1.2

rˆt
Thailand

2000

πt

7.5

5.0

0.1

1.6

1.1

1.4

1.8

π̂t(s)

9.5

5.6

2.7

2.9

1.7

1.8

1.7

π̂t(l)
rˆt

4.6

5.0

4.5

3.3

2.6

2.4

2.4

10.6

–1.8

–1.3

–1.1

0.5

–0.1

–0.5

NOTE: For the years surrounding the switch to IT, this table shows the inflation rate (πt ); expected inflation one year in the future
(π̂ t(s)); expected inflation six to ten years in the future (π̂ t(l)); and the ex-ante real interest rate, rˆt , measured as the policy rate less
π̂t(s). All variables are measured in the fourth quarter of the given year.

80

J U LY / A U G U S T 2 0 0 4

Commentary
Harald Uhlig

INTRODUCTION AND SUMMARY OF
THE PAPER

I

nflation targeting has become the new gospel
for conducting monetary policy. Its merits have
been stressed repeatedly by a large literature,
most recently by Michael Woodford (2004). The
appeal is obvious. Monetary economists have concluded that, by and large, monetary policy cannot
do much more than achieve some desired level of
inflation. Even to the extent that monetary policy
can have an influence on real activity, this lever
should be used only with great caution, as it may
endanger the reputation to effectively fight inflation over the longer-term horizon, if used opportunistically. The argument of Kydland and Prescott
(1977)—that monetary policy pursuing short-term
goals in a discretionary, opportunistic manner is
worse than a policy committed to and sticking to
an a priori well-chosen course of action—is fundamental and well understood. Inflation targeting is
often advertised as a way for monetary policy to
achieve an additional degree of this desirable commitment. Inflation targeting is furthermore sold as
an effective communications policy, as it makes
the public focus on inflation as the macroeconomic
variable (among those the public cares about) that
is most effectively influenceable by monetary policy.
In sum, inflation targeting is an appealing idea on
a priori theoretical grounds.
But inflation targeting has long moved beyond
the realm of academic conferences and scholarly
journals. It actually may have had its origins in the
practical choices undertaken by some central banks,
and it is now in use by a large variety of central
banks around the world. Thus, the interesting question is whether inflation targeting is also appealing
because of the experiences gained in practice. This
is, in essence, the question asked by Andrew Levin,
Fabio Natalucci, and Jeremy Piger (hereafter, LNP)
in their paper. Their paper follows several other,

similar papers in the recent literature, most notably
Ball and Sheridan (forthcoming). LNP investigate
the experience of inflation targeters and non-targeters since 1994 in a number of Organisation for
Economic Cooperation and Development (OECD)
countries. They also evaluate the experience with
inflation targeting in a number of emerging market
economies.
In contrast to Ball and Sheridan as well as several
other previous papers, LNP find a number of clear
differences between inflation targeters and nontargeters. They investigate inflation forecasts at a
number of horizons and find that the changes in
inflation forecasts are correlated with lagged averages
of inflation for non-targeters but uncorrelated for
inflation targeters. They find that inflation is more
persistent for non-targeters. While the difference is
rather small for inflation measured by the consumer
price index (CPI), it becomes substantial once the
attention is focused on core inflation, i.e., leaving
out prices for food and energy, thus refining the
results by Ball and Sheridan (forthcoming), who
focused only on CPI inflation. LNP document that
gross domestic product (GDP) growth volatility is the
same for targeters and non-targeters, but inflation
volatility is higher for inflation targeters, but argue
that this is due to experiencing larger shocks rather
than the monetary policy strategy itself. As for
emerging market economies, they document that
the introduction of the inflation target moves inflation expectations down gradually: The transition is
a smooth one and not characterized by a break at
the introduction of IT.
This is all fascinating. It is well-crafted research,
providing us with a wealth of additional insights and
differences between countries that have adopted
inflation targeting and those that have not. The
choice of empirical relations these authors look at—
for example, the variety of links between inflation
forecasts and inflation or the persistence of core
inflation—is well chosen in that these are some of

Harald Uhlig is a professor of economic policy at the Institute for Economic Policy, School of Business Administration and Economics, Humboldt
University.
Federal Reserve Bank of St. Louis Review, July/August 2004, 86(4), pp. 81-88.
© 2004, The Federal Reserve Bank of St. Louis.

J U LY / A U G U S T 2 0 0 4

81

REVIEW

Uhlig

the key dimensions where one would believe a priori
that inflation targeting may make a difference. The
choices are creative in that some of these dimensions
have not received sufficient attention before. The
documentation on the details of inflation targeting
in a number of emerging market economies will be
useful input into further research into these issues.

ASSESSMENT
I shall by and large not quarrel with the empirical
findings. The correlations found by the authors seem
to be there, and they can be read as a list of interesting differences between countries that have formally
adopted inflation targeting and those that have not.
There is an issue regarding the bias in their estimate
of the regression coefficient of inflation forecasts
on current inflation: However, this issue is not the
key concern I have about this paper, and I shall
discuss it last.
Rather, I shall quarrel with the interpretation of
these findings. The title of the paper already suggests
that the authors are not merely interested in uncovering correlations, but in documenting causation,
interpreting these numbers as macroeconomic
effects of inflation targeting. In short, they wish to
have a reader read these findings as an answer to
the policy question of interest: Does inflation targeting matter? And even if the authors abstained from
any interpretation in that direction, this would be
the question foremost on the mind of any reader of
their paper.
I am skeptical that such a causal interpretation
is legitimate. In particular, I am wondering about
four issues. First, which central bank should be
classified as inflation targeting? That is, is this classification useful? Second, has inflation targeting been
beneficial for the variables of ultimate interest, i.e.,
for reducing the level and volatility of inflation, for
reducing the volatility of output growth, and perhaps
for increasing the level of economic activity? Third,
the adoption of inflation targeting seems to me to
be largely a choice endogenously explained by some
of the variables under investigation: This is true in
the OECD sample, but it is particularly true for the
emerging market economies. For example, fiscal
restraint there probably matters far more than the
adoption of a formal inflation target. Finally, given
all the supposed benefits of inflation targeting, one
has to wonder what central banks did before inflation targeting and whether inflation targeting will
help prevent bad monetary policy in the future.
82

J U LY / A U G U S T 2 0 0 4

WHICH CENTRAL BANKS PURSUE
INFLATION TARGETING?
In his comments on the Ball-Sheridan paper,
Mark Gertler (forthcoming) pointed out that classifying countries in the run-up to the European monetary union as non-inflation targeters seems a bit
arbitrary in light of the fact that there were some
rather explicit preconditions in terms of achieved
inflation rates for entering the monetary union.
This applies to the LNP paper as well. More generally, where is the big difference between a central
bank that pursues the goal of price stability as a
prerequisite to monetary union and a central bank
that issues official statements about its price stability
goals? Do we really think that printing fan charts of
inflation in monthly reports can make a substantial
difference in monetary policy?
Take the case of the Bundesbank, a non-inflation
targeter according to the classification in LNP. Indeed,
officially, the Bundesbank has pursued money
growth rate targeting for many years. But, as has been
noted by a number of authors, the Bundesbank has
always chosen to ignore violations of the announced
money growth rate target, if it helped in pursuing
some other, more important goal—most notably
price stability. Secondly, the money growth rate target
has been derived from some underlying goal regarding the desired inflation rate, even if that may not
have been stated as explicitly as, say, by the Bank
of England. Finally, public debates about monetary
policy choices by the Bundesbank rarely evolved
around keeping or violating money growth rate targets: Rather, the debates were practically always in
terms of inflation.
The literature has drawn a distinction between
central banks that attach a high weight to inflation
versus central banks that pursue inflation targeting
(which is also consistent with a low weight on inflation in the objective function of the central bank);
it has also emphasized that the distinction between
targeters and non-targeters is a distinction regarding
their communication strategies. But if the communication, however done, ends up leading to a public
discussion and evaluation of monetary policy
choices regarding achieved levels of inflation and
implications for inflation rates in the future (as was
undoubtedly the case for the Bundesbank) and if
by contrast public debates focus on exchange rate
movements and the general state of the economy
rather than on keeping or violating some inflation

FEDERAL R ESERVE BANK OF ST. LOUIS

target (even if the appropriate fan charts are printed
in some monthly bulletin), isn’t it a bit artificial to
call such a bank an inflation targeter and the
Bundesbank a non-targeter? One could argue that
successful inflation targeting should precisely lead
to the end of the discussion about likely inflation
rates in the future and that the debate should therefore shift to other issues. But this misses the point.
Inflation targeting presumably works only if the
central bank is somehow held accountable for violating its inflation goals: If a reaction in the public
debate is unlikely even when violations occur, and
if the central bank that is officially targeting inflation
does not care either, then inflation targeting takes
place on paper only. Indeed, this is precisely the
argument that has been made regarding money
growth rate targeting for the Bundesbank: It applies
with similar force to inflation targeting.
Moreover, proponents of inflation targeting
typically argue for “flexible inflation targeting,”
which seems to mean that central banks also weigh
in objectives other than pursuing their inflation target when making their policy choices. For example,
a central bank targeting a particular rate of inflation
may decide to pursue a looser monetary policy in a
recession and a tighter monetary policy in a boom,
even if this leads to medium-term violations of
their inflation target, as long as this is all explained
well to the public. Interestingly, money growth rate
targeting can be viewed as accomplishing precisely
this. If a central bank pursues a particular rate of
money growth, calculated as the sum of desired
inflation and average economic growth and subtracting average changes in the velocity of money, then
money growth will be high; thus, monetary policy
will be loose precisely when economic growth falls
short of the average, and the other way around.
Money growth rate targeting has gone out of fashion,
but it almost certainly is a better way to implement
“flexible inflation targeting” than a policy that rigidly
enforces a particular inflation rate period by period!
So, perhaps, the Bundesbank should be classified
as a flexible inflation targeter, while other central
banks that rigidly focus on inflation in their reports
should be excluded from this category. Since the
issue at stake is whether flexible inflation targeting
should be used or not, the categorization by LNP
(using categorizations put forth by a number of previous authors, obviously, so they are not to blame)
could be entirely misleading.

Uhlig

HAS INFLATION TARGETING BEEN
BENEFICIAL?
Despite the arguments given above, I shall proceed with the categorization and the results in LNP
as the working hypothesis regarding the differences
between inflation targeters and non-targeters. The
key policy question in light of this experience is: Has
inflation targeting been beneficial? More precisely,
has inflation targeting been beneficial for the variables of ultimate interest, i.e., for reducing the level
and volatility of inflation, for reducing the volatility
of output growth, and perhaps for increasing the
level of economic activity? It may be nice (if true)
that long-run inflation expectations are anchored
more firmly for inflation-targeting countries, but
how helpful has that been for the variables we ultimately care about?
Here, the inflation targeters do not seem to fare
well. Table 4 in LNP reveals that the standard deviation of inflation for inflation targeters has been 1.54,
but only 0.81 for non-targeters, while output growth
volatility has been essentially the same (2.63 versus
2.48). Based on these numbers alone, one certainly
would not want to make the case that adopting
inflation targeting is a good idea.
The authors stress that this unconditional perspective is misleading because inflation-targeting
countries may have been hit by larger shocks or
have started from economic conditions that were
worse. Indeed, this argument is at the heart of the
Ball-Sheridan (forthcoming) paper: If one takes
initial conditions into account, most of the differences between inflation targeters and non-targeters
are explained by “regression to the mean.”
Would the authors here reach the same conclusion? That would then say only that there really is
no substantive difference in economic activity due
to the introduction of inflation targeting; and this
again does not provide an argument in favor of introducing IT (nor, obviously, an argument against it). It
seems to me that much more work than is in this
paper is required before it is possible to conclude
that inflation targeters have been more successful
in containing shocks hitting the economy than nontargeters have been. One could do this, it seems to
me, by more finely identifying the shocks hitting
these economies with, say, a multivariate VAR, and
to then assess their impact on monetary policy
choices as well as inflation expectations.
Examining Figure 1 in the LNP paper reveals
that inflation expectations gradually kept declining
J U LY / A U G U S T 2 0 0 4

83

REVIEW

Uhlig

for the non-targeters, but not so for the inflationtargeting countries. While one could debate whether
inflation expectations have declined too much in
Japan, overall this figure does indicate to me that
the non-targeters have been more successful in
bringing inflation down to the (currently) most
desired level, somewhere between 1 percent and 3
percent. Inflation targeting does not strike me as
virtuous monetary policy if the target is too high!
The analysis also reveals that the inflation forecasts
had similar volatility across both groups of countries:
While there may have been more external shocks
in the targeting countries, I am nonetheless surprised
that the targeters apparently were not able to offset
these shocks sufficiently to anchor long-term expectations more firmly.
Finally, the authors document that inflation
volatility has been of a more transitory rather than
persistent nature for targeters than for non-targeters.
In particular, the response of core CPI is more transitory in targeting countries. So, perhaps one could
make the case that inflation targeting leads to a
shift of inflation volatility from the low-frequency
spectrum to higher frequencies. But would that be
desirable? Probably not. Recent models of the New
Keynesian variety allow for some indexation of
otherwise sticky prices to past or to expected inflation: In these models, ongoing inflation or predictable inflation (or deflation, for that matter!) is not
particularly harmful. Instead, the economic distortions mainly come from distorting relative prices
between firms that can adjust their prices in response
to current shocks and firms that cannot. In short, in
these models, low-frequency volatility of inflation
is ok, but high-frequency volatility is bad for the
economy and leads to an overall lower level of
economic activity. If this is what would happen with
inflation targeting, which seems to be what the
empirical results suggest, then this would be an
argument against inflation targeting, not for it.

ENDOGENEITY
An overarching problem in moving from interpreting the correlations as indicating causation is
the endogeneity of the introduction of inflation
targeting. While this could help the case in favor of
inflation targeting (for example, if countries with
highly volatile inflation adopt inflation targeting,
one cannot blame that high volatility on inflation
targeting), one should not move to a causal interpretation, given the evidence currently presented.
Put differently, inflation targeting is probably often
84

J U LY / A U G U S T 2 0 0 4

introduced in the aftermath of some mild or strong
crisis or some general overhaul of institutional
structures.
The endogeneity issue is already important for
the OECD countries under consideration. For example, inflation targeting was introduced in the United
Kingdom alongside a whole set of institutional
changes, most notably a greater degree of independence for the Bank of England. Indeed, one can read
the Ball-Sheridan (forthcoming) paper as arguing
that inflation targeting was typically introduced
when the economic situation was sufficiently bad.
By contrast, things “look OK” in the United States
and the European Monetary Union. The need to
introduce inflation targeting there simply has not
been as pressing, so it has not been done.
The endogeneity issue is of even stronger force
for the emerging market economies investigated
by LNP. A number of recent papers in the literature
have documented that fiscal consolidation and
reform have been of greatest importance in allowing
monetary policy to pursue the goal of price stability.
As an example, the currency board in Argentina did
not break down because the central bank lacked
commitment, but rather because the fiscal situation
deteriorated. Likewise, the dramatically high inflation
rates in Russia at the beginning of the 90s were not
a choice by central bankers uninterested in the
pursuit of price stability, but rather were because
the only way to finance government expenditures
in the absence of a functioning tax system (aside
from borrowing) was seignorage.
These fiscal considerations are key in evaluating
the success or failure of monetary policy reforms in
emerging market economies. They are largely absent
in the paper at hand and should be investigated
seriously in future research following up on LNP.
The interesting question remains whether inflation targeting has contributed above and beyond
fiscal consolidation or general institutional reforms.
To answer this question, one would need to find a
clever instrument for the introduction of inflation
targeting, one uncorrelated with these other forces.
Finding a convincing instrumental variable seems
to be a thorny problem here, but one worthy of
attention. For example, one might try the fraction
of academics on the boards of central banks (who
may presumably be more inclined to move toward
the academically appealing idea of inflation targeting); but even this might just be a consequence of
general reforms. An alternative is to control more

FEDERAL R ESERVE BANK OF ST. LOUIS

Uhlig

Figure 1

Figure 2

Annual U.S. Inflation Rates

Inflation 10-Year Standard Deviation (U.S.)

Percent
15

Percent
6

10

5

Standard Deviation

4

5

3
19
33
19
38
19
4
19 3
48
19
53
19
58
19
63
19
68
19
73
19
78
19
83
19
88
19
93
19
98

0

2

–5

1

Year

carefully for variables indicating reforms, e.g., the
ratio of fiscal deficits to GDP and the like.
The authors are careful in not overstating their
results, and that is good. For example, they say that
“it is not completely obvious, however, the extent to
which [the reduction in inflation] can be credited...
to IT.” They observe that the reduction in inflation
expectations is gradual and that no sharp break can
be observed at the time of its introduction. This all
seems to me to be in line with the view that the
introduction of inflation targeting at some point is
simply a step that typically happens in countries
undergoing reform.

WHAT DID CENTRAL BANKS DO BEFORE
THERE WAS INFLATION TARGETING?
The paper investigates the episode since 1994
to evaluate the macroeconomic effects of inflation
targeting. That may be a rather short period to evaluate the success or failure of certain monetary policy
principles. Of course, inflation targeting did not exist
before that time (to my knowledge). So, for example,
what did the Fed do as a non-targeter (but perhaps
recently influenced by the targeting debates) before
inflation targeting was a policy option? Was monetary policy necessarily much worse?
Figure 1 shows inflation in the United States,
using annual data on the personal consumption
expenditures, taken from the national income and
product accounts (NIPA) tables published by the
Bureau of Economic Analysis. There were several
episodes of low inflation, like the 1930s, the 1950s

40

48
19
52
19
56
19
60
19
64
19
68
19
72
19
76
19
80
19
84
19
88
19
92
19
96
20
00

19

–15

44

0
19

InflationPCE

19

–10

Year

and 1960s, and the episode since the early 1980s.
But not all low-inflation episodes are equal. Figure 2
shows the evolution of the volatility of inflation,
calculated as the standard deviation of annual inflation over the preceding ten years. One can now see
more clearly than in Figure 1, that there have been
two episodes in which U.S. monetary policy has been
successful in stabilizing inflation. These standard
deviations are low in the 1960s as well as in the
1990s. Both episodes look remarkably similar in
that regard. Obviously, one needs to keep in mind
that these standard deviations are “backward looking,” i.e., a low standard deviation plotted for a particular year really means that inflation has been stable
in the preceding ten years up to and including the
year in question.
It is interesting to compare the inflation volatility
with the corresponding volatility in real output
growth (see Figures 3 and 4). While the literature
occasionally emphasizes a trade-off between these
volatilities (see, e.g., Uhlig, 2001), these figures show
substantial comovement between both volatilities
for the United States. These figures seem to suggest
that an environment of low and stable inflation
helps to reduce output volatility and support economic activity.
From that perspective, both the 1950s and
1960s as well as the 1980s and 1990s have been
particularly successful episodes of U.S. monetary
policy (or may have been episodes in which monetary policy was “lucky” due to the absence of major
disruptions such as oil price shocks). These successJ U LY / A U G U S T 2 0 0 4

85

REVIEW

Uhlig

Figure 3

Figure 4

Comparing 10-Year Standard Deviations
(U.S.)
Percent
12

Scatter Plot of 10-Year Standard Deviations:
Inflation Versus Real GDP Growth, U.S.,
1940-2002
10-year Standard Deviation of Real GDP Growth
12

10

Real GDP Growth Standard Deviation

10

Inflation Standard Deviation

8

8
6

6

4

4

2

2

0

0
45
19
50
19
55
19
60
19
65
19
70
19
75
19
80
19
85
19
90

0

19

35
19
40

19

19

30

Inflation, Growth
1

2

3

4

5

6

10-year Standard Deviation of Inflation

Year

ful episodes were achieved without an explicit
inflation-targeting regime. I doubt that introducing
an explicit inflation-targeting regime could have
produced better monetary policy during these episodes. The post-1994 comparison by LNP focuses
on an episode in U.S. monetary policy that already
was very successful: It is therefore not surprising
that one has a hard time interpreting the evidence
presented by LNP as making a strong enough case
for the introduction of inflation targeting in the
United States.
What is much more crucial, though, is the
question of how one can avoid an episode like the
1970s (or, even more dramatically, an episode like
the one before 1950), in which inflation volatility
and output volatility were both high. Can we really
be confident that political forces will continue to
appoint central bankers to the FOMC who understand the benefits of low and stable inflation, who
understand the importance of sticking to rules rather
than using discretion, and who continue to put these
principles into practice? As an “insurance,” wouldn’t
it be wonderful to somehow enshrine some of these
principles underlying the currently successful U.S.
monetary policy for the future? Here is where the
real issue lies. And here is where inflation targeting
can help. Once an explicit inflation target has been
announced and once the Fed explains its policy
choices in terms of this target, a repeat of the 1970s
or a repeat of the episode prior to the 1950s still
cannot be ruled out. But it will be more difficult,
86

J U LY / A U G U S T 2 0 0 4

since it would require a more clearly visible deviation from principles established before. These are
sunshine days for U.S. monetary policy, and there
is wide agreement among monetary policymakers
as well as academics as to the key goals of monetary
policies and the principles underlying good conduct.
Enshrining them now supports a longer life for this
type of monetary policy.

AN ESTIMATION BIAS
A key claim in the LNP paper is that inflation
forecasts are more highly correlated with past inflation in non-targeting countries than in inflationtargeting countries. The authors have also
documented, however, that inflation is more
volatile—in particular—at high frequencies, in inflation-targeting countries. This leads to a bias in their
coefficient estimate.
Consider their regression equation (1),
(1)

∆π̂ t( q ) = λ + β ∆ π t + εt ,

where πt is the inflation rate, π̂t(q) is the q-period-ahead
forecast of inflation, and ∆ denotes the change in
the variable in question. Suppose that inflation is
the sum of some persistent process ζt—a random
walk, say—plus i.i.d. noise υt ,

π t = ζt + υt .
Thus, ∆πt is a noisy signal about the change ∆ζt

FEDERAL R ESERVE BANK OF ST. LOUIS

of the inflation trend. Given some other source of
signal about the trend change as well as some prior
view, the best forecast will be some weighted average
of the recent change in inflation as well as the other
source of information and the prior, with the weights
proportional to the precision of the signal. Put differently, the larger the variance of the noise υt , the lower
the weight of ∆πt in the inflation forecast and thus
the lower the coefficient β. Econometrically, the
regressor ∆πt is a noisy version of the “true” regressor
∆ζt, leading to a downward bias in β, which is the
larger the larger is the variance in υt (i.e., the larger
is the high-frequency volatility in inflation). Since
the authors find higher high-frequency volatility in
inflation for inflation targeters, it is therefore not
surprising that the correlation of inflation forecasts
with current inflation is lower than for non-targeting
central banks.
Obviously, the other explanation for the lower
correlation is that the variance of the trend changes,
∆ζt, is lower or even zero for inflation targeters, as
they are fixing the desired level of inflation, whereas
it may be trending for non-targeters. This could and
should all be sorted out in a fully specified model,
including a signal-extraction-type equation for generating the forecast. The figures given in the paper
are suggestive that, indeed, desired inflation is subject to larger changes in the non-targeting countries,
as the authors suggest. As argued above, this does
not imply that inflation targeters have been more
successful with respect to the variables we care
about, namely, low and stable inflation. This economic issue is thus probably of greater relevance
than the econometric issue of the estimation bias.
Still, the bias deserves more attention and estimation of a more appropriate model.

CONCLUSIONS
This paper is interesting, creative, and informative. It deepens our knowledge about the differences
between countries that have adopted inflation targeting and those that have not. I do not believe that
the paper can empirically support the conclusion
that adopting an inflation target leads to more successful monetary policy; nor does it allow the opposite conclusion. More research is needed to get at
the issue of causation: This research seems feasible
to do and should be done.
What, then, is one to conclude? Should central
banks, specifically, the Federal Reserve or the
European Central Bank (ECB), adopt some version
of inflation targeting? It seems to me that the case

Uhlig

really rests in the power of the logic of the argument, not in the empirics presented by LNP. Both
the Federal Reserve and the ECB keep on emphasizing that keeping inflation low and stable is the best
contribution they can make to promote economic
activity. Furthermore, both central banks try to
minimize the distortions on financial markets
caused by monetary policy: They do so by communicating changes in monetary policy stances in
advance and by avoiding to “turn the wheels” too
fast. Keeping long-run inflation expectations stable
and reasonably low (but perhaps not too low) is of
great importance to both these central banks. Thus,
setting a clear goal for medium- and long-term
inflation and discussing the policy choices also,
although presumably not exclusively, in terms of
how quickly and in which way they will achieve
these goals can only help the Fed and the ECB to
pursue their desired policies yet more effectively.
Inflation targeting for these central banks would not
change their policies as they are currently pursued,
but rather would modify their existing communications. Inflation targeting will neither lead to dramatic
changes nor will it end the search for better monetary policy. Inflation targeting will not preclude the
discussion of other policy objectives for monetary
policy: Instead, it allows for a compartmentalization
and structuring of the arguments. Furthermore, it
may help to enshrine the current wisdom that low
and stable inflation is a key goal for monetary policy,
thus helping to avoid another return to inflationary
episodes like the 1970s and the corresponding distortions to economic activity. In sum, inflation targeting is an addition to current communications and
discussions about monetary policy and offers a
gradual improvement by helping to further organize
the internal as well as public debate on monetary
policy choices and by helping to commit both the
public and the central banks to keeping inflation
low and stable. Viewed this way, inflation targeting
is a good idea and should be adopted both by the
Fed and the ECB.

REFERENCES
Ball, Laurence and Sheridan, Niahm. “Does Inflation
Targeting Matter?” in Ben S. Bernanke and Michael
Woodford, eds., The Inflation Targeting Debate. Chicago:
University of Chicago Press (forthcoming).
Gertler, Mark. Comment on L. Ball and N. Sheridan, “Does
Inflation Targeting Matter?” in Ben S. Bernanke and

J U LY / A U G U S T 2 0 0 4

87

REVIEW

Uhlig

Michael Woodford, eds., The Inflation Targeting Debate.
Chicago: University of Chicago Press (forthcoming).
Kydland, Finn E. and Prescott, Edward C. “Rules Rather
than Discretion: The Inconsistency of Optimal Plans.”
Journal of Political Economy, June 1977, 85(3), pp. 473-91.
Levin, Andrew T.; Natalucci, Fabio M. 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.
Uhlig, Harald “The Role of National Central Banks and of
Different Policy Cultures,” in Charles Wyplosz, ed., The
Impact of EMU on Europe and the Developing Countries,
WIDER Studies in Development Economics. Oxford: Oxford
University Press, 2001, pp. 76-110
Woodford, Michael. “Inflation Targeting and Optimal
Monetary Policy,” Federal Reserve Bank of St. Louis
Review, July/August 2004, 86(4), pp. 15-41.

88

J U LY / A U G U S T 2 0 0 4

The Role of Policy Rules in Inflation Targeting
Kenneth N. Kuttner
No rule is so general, which admits not some
exception.
—Robert Burton, Anatomy of Melancholy

1. INTRODUCTION

M

ore than 13 years have elapsed since the
Reserve Bank of New Zealand’s pioneering
introduction of a formal inflation target
in 1990, a framework subsequently adopted by at
least 21 other central banks. Collectively, these 22
countries represent more than 132 country-years
of experience with inflation targeting (IT).1 This
accumulation of experience has led to a growing
understanding of the practical and institutional
features of the policy; see, for example, Bernanke
et al. (1999), Sterne (2002), Mishkin and SchmidtHebbel (2002), and Truman (2003). Roughly in
parallel with central banks’ embrace of IT, there
has been an explosion of research on monetary
policy rules—spawned in no small part by Taylor’s
(1993) influential paper. This line of research has
blossomed in recent years, especially with the
theoretical contributions of Clarida, Galí, and Gertler
(1999), Svensson (2003), Woodford (2003), and
Giannoni and Woodford (2003a,b) on optimal policy
rules, to name just a few.
Given these parallel developments in central
banking practice and monetary theory, it is no surprise that a great deal of recent research has modeled
IT as some sort of a monetary policy rule. Views differ
on the usefulness of describing IT in these terms,
however. Coming at the question from a practical
1

An annotated listing of inflation targeters can be found in Mishkin
and Schmidt-Hebbel (2002). A similar list of inflation targeters appears
in Truman (2003) and in each volume of the International Monetary
Fund’s International Financial Statistics.

standpoint, Bernanke et al. (1999) describe IT as a
“framework” rather than as a rule. In a similar vein,
Gavin (2004), characterizes IT as “management by
objective,” rather than as a well-defined policy rule.
On the other end of the spectrum, Svensson (1999)
defines IT as a monetary policy rule derived from an
explicit optimization problem. Much of the difficulty
in defining IT is due to its humble origins in central
banking practice, and policymakers’ pragmatic
search for a suitable nominal anchor. Svensson’s
definition can therefore be seen as part of a broader
effort to retrofit macroeconomic theory to a policy
that was developed largely in the absence of a formal
theoretical framework.2
Clearly, neither polar view can adequately
characterize IT as it is actually practiced. For IT as a
“framework” or a “management objective” to make
any difference to macroeconomic outcomes, it must
translate into some sort of change in central bank
behavior. And this change, in turn, should have
implications for the empirical policy rule used to
describe policymakers’ behavior. Similarly, surely
no central bank sees itself as an automaton mechanically implementing a policy rule. The goal of this
paper is therefore to determine where central banking practice lies on the “guidelines versus rules”
spectrum and, specifically, to assess empirically
the extent to which IT can be described in terms of
simple monetary policy rules.
The task of characterizing IT in terms of a policy
rule is complicated both by conflicting definitions
of the term “policy rule” and by differing interpretations of what IT entails in practice, however. In an
attempt to clarify what is meant by the term, section 2
2

As Goldfeld (1984) quipped in a different context, “An economist is
someone who sees something working in practice and asks whether
it would work in principle.”

Kenneth N. Kuttner is the Danforth-Lewis Professor of Economics at Oberlin College. This paper was written while the author was on the staff of
the Research Department of the Federal Reserve Bank of New York. The author thanks Marc Giannoni and Adam Posen for extremely helpful discussions, Özer Karagedikli, Monika Piazzesi, Christopher Plantier, Lars Svensson, Andrea Tambalotti, Dan Thornton, Mike Woodford, and Anders
Vredin for comments on earlier drafts, and Julia Schwenkenberg for research assistance. The author is especially indebted to Georgios Chortareas,
Özer Karagedikli, and Anders Vredin for making their respective central banks’ historical forecast data available.
Federal Reserve Bank of St. Louis Review, July/August 2004, 86(4), pp. 89-111.
© 2004, The Federal Reserve Bank of St. Louis.

J U LY / A U G U S T 2 0 0 4

89

REVIEW

Kuttner

reviews the various definitions of “policy rule.”
Section 3 describes how IT is implemented and how
this practice might be mapped into the sorts of rules
summarized in the preceding section. Section 4 then
characterizes the behavior of major IT central banks
econometrically, with the various interpretations
of “policy rule” as a guide. The novel feature of our
empirical analysis is the use of central banks’ own
inflation and output forecasts, exploiting the overlapping nature of these forecasts in order to estimate
the response of policy to new information. Section 5
concludes.

2. POLICY RULES: A GUIDE FOR THE
PERPLEXED
The explosion in research on monetary policy
rules has resulted in a proliferation of definitions
of the term “policy rule.” This has, not surprisingly,
led to some confusion in the literature, and discussions of the topic often flounder on issues of terminology. Naturally, the answer to the question of
whether IT can be described by a policy rule will
depend a great deal on what exactly is meant by
the term.3 This section provides a brief review of
some of the alternative definitions in an effort to
clarify some of these issues.

Conditional Versus Unconditional Rules
The easiest rules to enforce, of course, are simple
ones. And in the context of monetary policy rules,
the simplest sort of rule would be something like the
fixed money supply rule analyzed in Rogoff (1985)
and King (1997). Such a “non-contingent” or unconditional rule represents an inflexible commitment
to a nominal anchor and obviously prohibits any
sort of response to economic conditions. In certain
models, this kind of inflexibility can result from
assigning a zero weight to output fluctuations in the
central bank’s loss function, a case King (1997) refers
to as that of the “inflation nutter.” Such a rule would,
by construction, not allow the central bank to (optimally) offset the effects of supply shocks on output.
And as pointed out by Rogoff (1985), such a rule
would also prevent the central bank from accommodating nominal money demand shocks, potentially
introducing more volatility in output and inflation.
It is this sort of inflexible rule that was criticized by
Friedman and Kuttner (1996), among others. Recent
3

Bofinger (2000) also grapples with the problem of mapping IT into
alternative definitions of a policy rule.

90

J U LY / A U G U S T 2 0 0 4

research has deemphasized these unconditional
policy rules, however, and the focus now is much
more on the design of flexible, conditional rules that
allow the policymaker to respond in a reasonable
(or even optimal) manner to economic conditions.

Ad Hoc Versus Optimal Rules
Among those “state contingent” rules that allow
policy to respond to economic conditions, a broad
distinction can be drawn between ad hoc policy rules
relating the policy instrument to some selection of
macroeconomic variables and those rules derived
from an explicit optimization problem. Taylor’s
(1993) eponymous rule
(1)

it=r*+π*+0.5 xt+1.5(πt – π*)

is of course the best-known example of the former;
here, it is the nominal policy interest rate, r* is the
equilibrium real rate of interest, π* is the desired
or “target” level of inflation, πt is the current rate
of inflation, and xt is the output gap.4 Another
example is the inflation forecast–based (IFB) rule
proposed by Batini and Haldane (1999), in which
the nominal interest rate depends on a distributed
lead of τ-period-ahead inflation forecasts made at
time t, πt+τ ,t (ignoring interest rate smoothing for
simplicity):
(2)

it = r * +π * + ∑ θτ π t + τ , t .
τ

By contrast, rules based on an explicit optimization problem are almost invariably based on setting
policy in such a way as to minimize a loss function
of the form
`

(3)

Et ∑ δ τ [(π t + τ − π*)2 + λ xt2+ τ ] .
τ =0

The optimal response of the interest rate derived
from (3) will in general not be given by (1) or (2),
although it is often possible to reverse-engineer
objective functions that would rationalize such rules.
Nonetheless, these sorts of rules tend to produce
reasonable (if not optimal) policy responses, provided
the sum of the inflation coefficients exceeds unity,
thus satisfying the Taylor principle. This observation
has led to a vast literature comparing the performance of simple ad hoc rules, like Taylor’s, with the
optimal policy response for a variety of models. See,
4

To make the rule operational, Taylor used the current year-over-year
inflation rate for πt and linearly detrended output for xt.

FEDERAL R ESERVE BANK OF ST. LOUIS

for example, Rudebusch and Svensson (1999) and
Williams (2003).

Targeting Versus Instrument Rules
Probably the most familiar way to characterize
the conduct of monetary policy is in terms of an
instrument rule involving the policy instrument
itself. Equations (1) and (2) are examples in which
the nominal interest rate is the policy instrument.
Although these examples are ad hoc rules, it would
also be possible to insert the relevant structural
relationships into the first-order condition from minimizing (3) to yield the optimal instrument rule for
a particular model. A distinction is also sometimes
made between explicit instrument rules involving
predetermined macro variables (i.e., ones that do not
depend on the current setting of the policy instrument) and implicit rules whose arguments are jointly
determined with the policy instrument (as would
typically be the case if the rule’s arguments included
forecasts). An explicit rule can be interpreted as a
simple “recipe” for policy, while implementing an
implicit rule (such as the IFB) requires that the policymaker account for the feedback between the instrument and the arguments of the rule.
A less familiar way to characterize policy is
directly in terms of the targeted variables themselves,
yielding what is referred to as a targeting rule. An
optimal targeting rule could either take the form of
simply pledging to minimize (3) (a “general” targeting
rule in Svensson’s terminology) or setting policy in
such a way as to satisfy the first-order conditions
from (3) describing the marginal trade-off facing the
policymaker between output and inflation stabilization (a “specific” targeting rule). With a quadratic
objective function, the first-order conditions imply
a linear marginal trade-off between inflation and
the output gap, such as those discussed in section
4. There are also ad hoc targeting rules involving
only the central bank’s goal variables, but which
are not explicitly derived from an optimization
problem. A prescription to achieve the desired
inflation rate at a fixed horizon could be classified
as a targeting rule of this type.
Because instrument rules can be derived from
targeting rules (and vice versa), the distinction
between the two is at some level artificial. The distinction is further blurred when a term involving
the change in the interest rate is included in the
objective function, presumably reflecting a preference for interest rate smoothing as in Giannoni and
Woodford (2003a,b). In this case, the policy instru-

Kuttner

ment would appear in the first-order condition,
making the targeting rule indistinguishable from
an implicit instrument rule.
Nonetheless, Svensson (2003) contends that
formulating policy in terms of a targeting rule has
several compelling advantages over an instrument
rule. Chief among these is that targeting rules can
more readily accommodate central bankers’ “judgment” regarding special factors affecting inflation
and output (macro models’ ubiquitous “add factors”),
as well as any unusual circumstances affecting the
efficacy of the monetary transmission mechanism.
Svensson shows that variables representing policymakers’ judgment will typically enter the instrument
rule formulation, rendering the implementation of
policy by that route more complex and less intelligible. By contrast, these judgment variables do not
appear in targeting rules. Intuitively, this is because
the targeting rule embodies the central bank’s imperative simply to do whatever it takes to achieve the
appropriate balance between output and inflation
stabilization, incorporating any judgment that might
be appropriate in selecting the instrument setting
needed to achieve the desired objective.

Rules Describing Discretion Outcomes
Versus Rules Derived from a
Commitment
Contributing to the terminological haze is the
fact that the term “policy rule” has been used to
describe both the outcome of discretionary policy
setting (i.e., re-optimizing each period) and a condition that commits the central bank to potentially
time-inconsistent actions in the future. In fact, as
shown by Clarida, Galí, and Gertler (1999), a Taylorlike instrument rule can be derived as the optimal
instrument rule for a central bank acting under
discretion. This observation means it is difficult to
discern in practice whether a central bank has acted
in a purely discretionary fashion or tempered that
discretion with some sort of commitment. Merely
finding that monetary policy is well-described econometrically by a simple policy rule like Taylor’s
therefore does not necessarily imply any sort of
commitment. This is not to say that optimal policy
rules derived under discretion are the same as those
derived under precommitment. Indeed, a generic
feature of the latter is some degree of “history
dependence”—i.e., for the policy response to depend
not only on the current state of the economy, but
the lagged state of the economy as well (typically,
J U LY / A U G U S T 2 0 0 4

91

REVIEW

Kuttner

the lagged output gap).5 As a practical matter, however, it may be difficult to detect this sort of behavior,
especially in the presence of an interest rate smoothing objective.

Mechanical Rules Versus Rules as
“Guidelines”
Related to the discretion-versus-commitment
dimension is the distinction between a rule that functions as a strict prescription for policy actions, versus
one that serves as a looser “guideline” or “framework” for the conduct of monetary policy. This
distinction has a long history. Simons (1936), for
instance, interpreted a policy rule as completely
precluding any intervention on the part of the
“authorities.” “Monetary rules,” Simons wrote,
“must be compatible with the reasonably smooth
working of the system. Once established, however,
they should work mechanically, with the chips
falling where they may” (Simons, 1936, pp. 13-14).
Clearly, it is this strict, mechanical interpretation
that Bernanke et al. (1999) object to as a characterization of IT.6
It is also apparent that this is not what Taylor
had in mind when he proposed his rule. “Policy
rules are frequently written down in the form of a
‘mechanical-looking’ algebraic formula…But this
does not mean that the only way that monetary
policy rules can be used is for the central bank to
follow them mechanically,” Taylor wrote. “On the
contrary, most recent proposals for monetary policy
rules assume that they would be used as guidelines
for policymakers, recognizing the need for some
discretion in using the rule” (Taylor, 2000, p. 209).
Svensson’s interpretation of a rule appears to
be closer to that of Simon’s than to Taylor’s. He finds
Taylor’s looser interpretation of a rule “not sufficiently specific to be operational” on the grounds
that “there are no rules for when deviations from
the instrument rule are appropriate,” which, he
argues, creates an “inherent lack of transparency”
(Svensson, 2003, p. 445). But Svensson does allow
discretion to enter through the arguments of the
5

Confusing matters still further: Considering only rules that depend
on the current state of the economy, the optimal instrument rule
under commitment is the same as that obtained under discretion
with a smaller weight on output in the loss function (λ ). See Clarida,
Galí, and Gertler (1999).

6

Even while arguing that a monetary rule should be mechanical,
Simons (1936) recognized that an unconditional rule, such as a fixed
supply of money, would probably be too inflexible.

92

J U LY / A U G U S T 2 0 0 4

policy rules: e.g., via judgmental adjustments to
forecasts or the policymakers’ assessment of the
policy setting required to attain the optimal outcome.

3. WHAT DOES INFLATION TARGETING
HAVE TO DO WITH POLICY RULES?
The goal of this section is to make a connection
between the various definitions of a policy rule discussed above and IT as it is actually practiced. To
that end, it first outlines three ways in which IT could
be interpreted within the context of monetary macro
models. Second, it summarizes the key features of IT
as it is practiced by self-described inflation targeters.
The section concludes by highlighting dimensions
along which the practice of IT does (or does not) map
into the theoretical characterizations of IT.

A Theoretical Taxonomy
Perhaps the weakest definition of IT is having
some desired rate of inflation, π*, (which need not
be announced) and employing a reaction function
or instrument rule that satisfies the “Taylor principle”
of a greater-than one-for-one response to expected
inflation, thus ensuring that eventually inflation
returns to π*. The reaction function may of course
also include a response to the output gap. This could
be called “weak form” IT. The reaction function need
not be optimal, in the sense of being derived from
an explicit loss minimization problem. Under this
definition, even following a simple ad hoc Taylorstyle rule will work. Galí (2002) and McCallum
(2002), among others, define IT in this way.
A stronger definition of IT (“semi-strong form IT”)
restricts membership in the club to those central
banks following an optimal monetary policy, i.e.,
setting policy in such a way as to minimize a relatively explicit loss function like (1) above. This is the
definition in Svensson (1999).7 This optimization
need not be carried out subject to any sort of precommitment, however. Indeed, the optimal instrument rule derived under the assumption of
discretionary, period-by-period optimization in
Clarida, Galí, and Gertler (1999) is described as a
rule that characterizes IT.
The strongest conceptual definition of IT
(“strong-form IT”) is in terms of optimal monetary
policy under conditions of precommitment (or alter7

Svensson’s (1999) definition goes even farther, defining IT narrowly
as an optimal targeting rule, although it is not clear why optimal policy
could not also be implemented with an instrument rule.

FEDERAL R ESERVE BANK OF ST. LOUIS

natively the “timeless” perspective). The optimal
policy rule derived under these assumptions involves
a commitment to a time-inconsistent course of future
action on the part of the central bank: e.g., reducing
inflation today by the promise to run negative output gaps in the future.8 Svensson (1999) seems to be
agnostic as to whether IT necessarily involves such
a precommitment, although he argues that IT can,
at least, help reduce or eliminate any inflation bias
resulting from an above-equilibrium output target.
The precommitment solution is clearly what King
(1997) had in mind in his description of IT, however,
and this is also apparently the view of Giannoni
and Woodford (2003c).9

Inflation Targeting in Practice
Although the practice of IT is anything but uniform across countries, a number of elements are
common to most self-declared inflation targeters.
These include10
• An emphasis on long-run price stability as
the principal goal of monetary policy: This
is not to say that price stability is the only
goal, however—only that other objectives
can be pursued only to the extent that they
are compatible with the inflation target.11
• An explicit numerical target for inflation and
a timetable for reaching that target: Most are
in the neighborhood of 2 percent, and all aim
to achieve the target at a horizon of no more
than two years.
• A high degree of transparency with regard to
monetary policy formulation: Most inflation
targeters publish a detailed report on general
economic conditions and the outlook for
inflation in particular, typically at a quarterly
frequency. In some cases, these reports
include numerical projections of key macroeconomic variables.
• A mechanism for accountability: Besides
promoting transparency, the central banks’
8

Or, in the case of Japan, increasing inflation today by the promise of
running positive output gaps in the future, as in Eggertsson and
Woodford (2003).

9

Oddly, the empirical application in Giannoni and Woodford (2003c)
is to the Federal Reserve, which is generally not considered a fullfledged inflation targeter.

10

A similar list of features is given in Truman (2003).

11

Debelle (2003) points out that Australia is something of an exception
in this regard, in that the Reserve Bank of Australia operates under a
dual mandate of price stability and full employment.

Kuttner

published inflation reports also provide a
means for ex post evaluation of inflation
performance. Failure to fulfill the inflation
target may require the central bank to take
specific steps. For example, should inflation
deviate by more than 1 percentage point from
its 2.5 percent target, the Governor of the Bank
of England is obliged to submit an open letter
to the Chancellor of the Exchequer explaining
the reason for the deviation and presenting
a timetable for a return to the target.
Although most inflation targeters share these
four broad features, there is considerable variation
in some of the specifics. At the level of implementation, for example, central banks differ with respect
to choice of price index (overall versus “core” inflation) and the particular form of the target (point
versus range). At a more substantive level, inflation
targeters also differ as to “goal independence”: those
that are only “instrument independent” pursue a
target set for them by the elected government, while
those that are also “goal independent” set their own
targets. There is also a great deal of heterogeneity
with respect to exactly what central banks communicate—and how; these differences are cataloged
in great detail by Fracasso, Genberg, and Wyplosz
(2003).
It is also interesting to distinguish full-fledged
inflation targeters from “near-inflation targeters”
on the basis of these features. For example, some
central banks—particularly those of emergingmarket or post-communist countries—have set up
most of the mechanics of IT, but lack the institutional means to make a long-run commitment to
price stability. Carare and Stone (2003) and Stone
(2003) refer to this regime as “Inflation Targeting
Lite.” Other central banks seem to have internalized
the price stability objective without adopting all the
trappings of outright IT. The Federal Reserve might
be put into this category. Although it also operates
under a dual mandate, in recent years price stability
has been increasingly emphasized as the primary
long-run goal of monetary policy.12 Nonetheless,
the Federal Reserve still lacks an explicit numerical
inflation target, and as discussed in Kuttner and
Posen (2001), its monetary policy formulation
12

In 1988, Greenspan stated: “We should not be satisfied unless the U.S.
economy is operating at high employment with a sustainable external
position and above all stable prices...By price stability, I mean a situation
in which households and businesses in making their saving and investment decisions can safely ignore the possibility of sustained, generalized
price increases or decreases” [Greenspan (1988), emphasis added].

J U LY / A U G U S T 2 0 0 4

93

REVIEW

Kuttner

remains considerably more opaque than that of the
typical inflation targeter.13

Will the Real Inflation Targeter Please
Step Forward?
The question now is how (and indeed whether)
the central bank practices described above map into
the various definitions of IT that have been proposed
in the theoretical literature. Based on the institutional
arrangements, which central banks qualify for
membership in the IT club? And to what extent
does this depend on the theoretical definition of IT
one has in mind?
“Weak form” IT is so weak as to be almost vacuous. Under this definition, virtually every major
central bank qualifies as an inflation targeter. It
implies merely that the central bank acts in such a
way as to eventually bring inflation back to its target,
and says nothing about the nature of the policy used
to get there. Thus, it would seem that the only
requirement for entry into the IT club is having a
reaction function coefficient of at least a certain size.
And indeed since this condition is also a requirement for a well-behaved solution to conventional
macro models, any country with a mean-reverting
inflation rate would ipso facto qualify as an inflation
targeter; only those with “unstable” inflation would
fail the test.14
It is hard to find institutional evidence that any
central bank conforms to either the “strong” or
“semi-strong” forms of IT. The problem is that no
central bank—inflation targeter or otherwise—
currently conducts policy with reference to an
explicit, publicly announced loss function or firstorder condition. Nor is there any documentary evidence to suggest that policymakers choose between
different policy options on the basis of any numerical estimate of the estimated “loss” associated with
the various options. Thus, on the basis of Svensson’s
definition of IT as an optimal targeting rule, it would
seem that no central bank qualifies as a bona fide
inflation targeter.15 Furthermore, even among self13

It is worth making a distinction between transparency in policy formulation versus transparency in implementation. The Federal Reserve has,
of course, become much more transparent in the implementation of
policy, especially since the practice of announcing changes in the
funds rate target began in February 1994.

14

In backward-looking models, failing to satisfy the Taylor principle
typically results in explosive inflation; in forward-looking models, it
can generate indeterminacies.

15

In the spirit of Friedman (1953), it might be argued that central bankers
act “as if” they were minimizing a loss function, even if they were not
consciously aware that fact.

94

J U LY / A U G U S T 2 0 0 4

proclaimed inflation targeters, only the Reserve
Bank of New Zealand (RBNZ) publishes a forecast
of the output gap, which is a key ingredient in conventional targeting rules.16
While no inflation targeter currently communicates in the language of optimization, all inflation
targeters talk a great deal. Indeed, as noted above,
communication is a central element of the practice
of IT. Formalizing the role of this IT-induced transparency has presented a theoretical challenge, however. From an optimal control standpoint, it makes
no difference whether the central bank keeps its
first-order conditions to itself or explains them in
detail to the public four times per year. Optimal
policy will be the same either way.
One hypothesis is that transparency and
accountability somehow allow the central bank to
overcome the time consistency problem inherent
in optimal monetary policy. One way to formalize
this idea is to assume private information on the part
of the central bank regarding its output or inflation
objectives. Herrendorf (1998) and Faust and Svensson
(2001) take this approach, as does Geraats (2002),
explicitly in the context of the decision to publish
a forecast. In Drazen and Masson (1994) and Agénor
and Masson (1999), the private information has to
do with the central bank’s preferences: specifically,
whether it is “strong” in the sense of assigning a low
weight on output fluctuations in its loss function
relative to a “weak” central bank. In the context of
both sets of models, transparency plays a direct role
in helping to reveal the unobserved private information. Less formally, King (1997) has argued that
transparency, accountability, and a clearly defined
objective all enhance central bank “credibility,”
defined as the ability to convince the private sector
that it will carry out policies that may be time inconsistent, and thus implement the optimal statecontingent rule.

4. ESTIMATED POLICY RULES FOR
THREE INFLATION TARGETERS AND
THE FED
We now turn to the question of how well the
behavior of IT central banks can be characterized
by simple policy rules of the sort discussed earlier
in section 2. We first consider conventional instrument rules relating the relevant short-term interest
16

Svensson (2000) contends this is an essential requirement for IT:
“policy decisions are consistently motivated with reference to published inflation and output(-gap) forecasts.”

FEDERAL R ESERVE BANK OF ST. LOUIS

rate to macroeconomic objectives, i.e., output and
inflation. Specific targeting rules—that is, equations
describing the optimal trade-off between output
and inflation—are taken up later in the section.
Ideally, one would want to use these estimates
to say which of the definitions of IT from section 3
best described the central banks’ behavior. This is
hard to do econometrically, unfortunately. Estimates
of instrument rules or reaction functions may reveal
something about whether the Taylor principle is satisfied, but without a fully and correctly specified
macro model, it is difficult (if not impossible) to distinguish between optimal and ad hoc behavior on
the part of the central bank. Similarly, the distinction
between discretionary and precommitment-based
policy is a very subtle one to discern empirically,
although Kuttner and Posen (1999) suggest that
precommitment-like behavior may explain the
observed decline in inflation persistence among
inflation targeters. At the very least, however, such
estimates can provide some information as to how
closely banks’ behavior conforms to simple rules
of one form or another.
This is, of course, not the first effort to describe
central banks’ behavior in terms of simple policy
rules. What distinguishes this paper from much of
the other work in the area is its use of the central
banks’ own published inflation and output forecasts,
rather than econometric proxies for the relevant
expectations.17 This approach has a number of
compelling advantages. First, it greatly simplifies
the econometrics, reducing the data requirements
and obviating the need for the two-stage GMM
method used by Clarida, Galí, and Gertler (2000).
Second, the central banks’ own forecasts are likely
to be more reliable than those based on simple
econometric methods.18 One reason is that the
central banks’ forecasts undoubtedly incorporate
a great deal of information not in the macro time
series, as well as informal “judgment” as to the most
likely outcomes. It is this kind of an information
17

Jansson and Vredin (2003) and Berg, Jansson, and Vredin (2002) use
Riksbank forecasts in their analysis, and Huang, Margaritis, and Mayes
(2001) utilize the RBNZ’s published projections. For the United States,
McNees (1986, 1992) uses the Federal Reserve’s internal Green Book
forecasts in estimating a forward-looking reaction function, an
approach that has been adopted by Orphanides (2001) and Boivin
(2003), among others.

18

All three central banks regularly assess their own forecasting performance: see, for example, McCaw and Ranchhod (2002) for New Zealand,
Pagan (2003) (and the references cited therein) for the Bank of England,
and the “Materials for Assessing Monetary Policy” appendix to the
Riksbank’s March Inflation Report.

Kuttner

advantage that Romer and Romer (2000) suggest
accounts for the superior performance of the Federal
Reserve’s Green Book forecasts. Finally, the published
forecasts presumably embody appropriate assumptions about the central banks’ intended policy
actions. Naive, unconditional econometric forecasts,
on the other hand, may imply a path of policy at
odds with central banks’ intentions.
The analysis in this section focuses on three
central banks: the RBNZ, the Bank of England, and
Sweden’s Riksbank. These three were chosen
because they are the three inflation targeters with
the longest track record of published, quantitative
forecasts. Other seasoned inflation targeters, such
as the Reserve Bank of Australia and the Bank of
Canada, make use of detailed projections internally,
but these projections have not been made available
publicly. Instead, their published reports contain
only more qualitative, impressionistic assessments
of the outlook for inflation and output, which are
less amenable to quantitative analysis.
Although all three of the central banks under
study report some sort of a forecast every quarter,
the nature of these forecasts differs considerably
across banks. Since 1997, the Bank of England’s
Inflation Report has consistently reported projections
for four-quarter inflation and real gross domestic
product (GDP) growth (mean, median, mode, and a
measure of uncertainty) for an eight-quarter forecast
horizon. The Riksbank, since 1992, has published
forecasts for Q4/Q4 real GDP growth and December/
December consumer price index (CPI) inflation for
the current year and the two following “out” years.
The RBNZ reports projections over a longer forecast
horizon than the other central banks, with some
forecasts going out as long as 15 quarters. The RBNZ
is also unique in that it is the only central bank ever
to have published estimates and forecasts of the
output gap—a key ingredient in typical targeting
and instrument rules.
The three central banks do differ somewhat as
to how they express their inflation targets. Over the
period covered by the analysis in this paper, the
Bank of England had a point target of 2.5 percent
for the retail price index less mortgage-related items
(known as the RPIX). The RBNZ’s current Policy
Targets Agreement (PTA) currently specifies a 1 to 3
percent range for overall CPI inflation. (From 1997
through 2002:Q3, the target range was 0 to 3 percent
for core CPIX inflation.) The Sveriges Riksbank has
a point target of 2 percent for overall CPI inflation,
with a ±1 percent tolerance around that target.
J U LY / A U G U S T 2 0 0 4

95

REVIEW

Kuttner

The policy assumptions underlying the three
central banks’ forecasts also differ on one key dimension. The Bank of England and Riksbank both report
constant-interest-rate forecasts, in which the interest
rate set by policy is assumed to remain unchanged
over the forecast horizon. Leitemo (2003) has interpreted this as a policy rule in which the current
interest rate is set at a level consistent with attaining
the inflation target.19 Vredin (2003) contends that
interpretation is inconsistent with the Riksbank’s
policy; indeed, its Inflation Report has at times
explicitly acknowledged that a significant divergence
between the forecast and the target would require
a change in policy. As discussed below, however, the
Bank of England’s and the Riksbank’s two-yearahead forecasts come reasonably close to their stated
targets, at least since 1997. This observation, along
with the well-known instability problems associated
with constant-interest-rate rules, led Vredin (2003)
to question whether these central banks’ forecasts
represent true constant-interest-rate forecasts.20
Unlike other inflation targeters, the RBNZ conditions
its forecasts on the time-varying interest rate given
by the inflation-forecast-based rule used in the staff
macro model.21 In doing so, the RBNZ avoids the
difficult methodological issues associated with
constant-interest-rate projections and provides an
outlook for interest rates and the economy that is
arguably more coherent than those of the other
central banks.
There are also subtle differences in the timing
of the forecasts relative to the policy decisions. The
Riksbank and the RBNZ both publish their forecasts
on the same day on which interest rate decisions
are made. Consequently, these two central banks’
forecasts are conditioned almost by construction
(given the time required to put together a new forecast) on the preceding period’s interest rate. By contrast, the Bank of England publishes its inflation
report two weeks after a Monetary Policy Committee
meeting, and as a result its forecasts are conditioned
19

Leitemo (2003) points out that such a constant-interest-rate rule is
time inconsistent, as the interest rate required to achieve the target
would be expected to change as the horizon advances, and derives a
method for calculating the time-consistent equilibrium under such a
constant interest rate rule.

20

Noting that the Riksbank’s two-year inflation forecast, since 1997,
has never fallen outside of the 1 to 3 percent range, Leeper (2003)
also questions the credibility of the Riksbank’s constant-interest-rate
assumption.

21

Details of the RBNZ’s IFB rule can be found in the appendix to Drew
and Plantier (2000).

96

J U LY / A U G U S T 2 0 0 4

on the current period’s policy action. Thus, the
Riksbank and RBNZ forecasts can be taken as predetermined with respect to the current-period interest rate, while those of the Bank of England cannot.
It is interesting to note that an explicit reaction
function or instrument rule does not figure prominently in any of the three central banks’ official
publications, which tend to focus instead on the
outlook for inflation, relative to its target, and a discussion of economic conditions more generally.
This is not to say that instrument rules have been
entirely ignored, however. Since 2000, the March
issue of the Riksbank’s Inflation Report has included
an assessment of monetary policy using an econometrically estimated “rule of thumb” based on the
Riksbank’s own inflation forecasts.22 According to
Archer (2003), a Taylor-style rule is used internally
at the RBNZ for assessing various policy options,
and one (the May 2001) issue of the Monetary Policy
Statement actually included such a rule-based
assessment. Nikolov (2002) reports that Bank of
England staff and the Monetary Policy Committee
periodically review the implications of a variety of
policy rules, although the output gap data used to
implement these rules are not made public.
For comparison, we also include an analysis of
Federal Reserve behavior based on the internal forecasts contained in the unpublished “Green Books,”
which are made public after a five-year lag.23 This
comparison is inexact at best, however, because
the policy assumptions underlying the Green Book
forecasts differ fundamentally from those embodied
in the inflation targeters’ forecasts. In particular,
while some plausible path for the funds rate target
is assumed in the Green Book, this policy does not
necessarily correspond to the FOMC’s intentions;
nor is it generally the case over the period analyzed
that the policy brings inflation back to a fixed target
by the end of the forecast horizon.24
22

The research underlying the rule of thumb described in the Inflation
Report can be found in Jansson and Vredin (2003) and Berg, Jansson,
and Vredin (2002).

23

Pre-1997 Green Book data can be found at
www.phil.frb.org/econ/forecast/greenbookdatasets.html.

24

The baseline assumption is typically (though not always) a constant
nominal funds rate; see Reifschneider, Stockton, and Wilcox (1997).
Another possibility would be to use the biannually published “central
tendency” forecast of the FOMC. Although this may conform to the
FOMC’s intentions more closely than the Green Book forecasts, the
conditioning assumptions (inflation objective, interest rate path) are
no clearer.

FEDERAL R ESERVE BANK OF ST. LOUIS

Kuttner

Figure 1
Inflation Forecasts and Targets
4
3.5
3

RBNZ Inflation Forecasts

5

3-Quarter Forecasts
8-Quarter Forecasts
Midpoint of Target Range
Current Inflation

Riksbank Inflation Forecasts

4
3

2.5
2
2
1

1.5

0

–2
M

ar
Se 94
p
M 94
ar
Se 95
p
M 95
ar
Se 96
p
M 96
ar
Se 97
p
M 97
ar
Se 98
p
M 98
ar
Se 99
p
M 99
ar
Se 00
p
M 00
ar
Se 01
p
M 01
ar
Se 02
p
M 02
ar
03

–1

Se
p
D 97
ec
M 97
ar
Ju 98
n
Se 98
p
D 98
e
M c 98
ar
Ju 99
n
Se 99
p
D 99
ec
M 99
ar
Ju 00
n
Se 00
p
D 00
e
M c 00
ar
Ju 01
n
Se 01
p
D 01
e
M c 01
ar
Ju 02
n
Se 02
p
D 02
ec
M 02
ar
Ju 03
n
03

0.5

3.5

4-Quarter Forecasts
8-Quarter Forecasts
Target
Current Inflation

0

1

Bank of England Inflation Forecasts

6

Green Book Inflation Forecasts

5.5
3

5
4.5

2.5

4
3.5

2

1.5

3
3-Quarter Forecasts
8-Quarter Forecasts
Target
Current Inflation

2.5
2
1.5

Properties of the Inflation Forecasts
We begin by simply examining the properties
of the central banks’ inflation forecasts. Figure 1
displays time series of the medium- and long-run
forecasts, along with the current year-over-year
inflation rate and (except for the United States) the
inflation target. The figure reveals considerable
variation in the relatively short-run three-quarterahead forecasts. Not surprisingly, the eight-quarterahead forecasts tend to fall much closer to the target.
Sweden in the early- to mid-1990s is something of
an exception, however, with even the two-year-ahead
forecasts well above the target until 1996.25 The Fed’s
forecast, shown in the lower right-hand panel, also

ar
J 8
N ul 87
o
M v 87
ar 7
J 8
N ul 88
ov 8
M 8
ar 8
J 8
N ul 89
ov 9
M 8
ar 9
J 9
N ul 90
ov 0
M 9
ar 0
J 9
N ul 91
ov 1
M 9
ar 1
J 9
N ul 92
ov 2
M 9
ar 2
J 9
N ul 93
ov 3
M 9
ar 3
J 9
N ul 94
o
M v 94
ar 4
Ju 95
N l9
o
M v 95
ar 5
Ju 96
N l9
ov 6
96

1
M

Se
p
D 97
ec
M 97
ar
Ju 98
n
Se 98
p
D 98
e
M c 98
a
Ju r 99
n
Se 99
p
D 99
e
M c 99
ar
Ju 00
n
Se 00
p
D 00
e
M c 00
ar
Ju 01
n
Se 01
p
D 01
e
M c 01
ar
Ju 02
n
Se 02
p
D 02
ec
M 02
ar
Ju 03
n
03

1

exhibits a great deal of variation, which reached 5
percent in 1989 before falling to less than 2 percent
with the disinflation that accompanied the 1991-92
recession.26
Another way to look at the same set of data is
to plot the forecast change in the inflation rate against
the deviation of current inflation from its target.
Forecasts that implied a return to the target within
the specified horizon would fall along a –45 degree
line; for example, inflation 1 percent above target
would imply a –1 percent forecast change. The eightquarter-ahead forecasts for the RBNZ fall reasonably
close to this –45 degree line, as shown in the upper
left-hand panel of Figure 2. For the United Kingdom,
26

25

According to Berg, Jansson, and Vredin (2002), this early stage of IT
was one in which the Riksbank was still struggling to establish the
credibility of its target.

7-Quarter Forecast
Current Core CPI Inflation

The Green Book’s forecast horizon over the 1986-96 period varied
between five and nine quarters, which is somewhat shorter than that
of the inflation targeters. A seven-quarter-ahead horizon was therefore
chosen so as not to lose too many observations.

J U LY / A U G U S T 2 0 0 4

97

REVIEW

Kuttner

Figure 2
Target Reversion of Inflation Forecasts
RBNZ Inflatio F rec s

Riksban I fl t o F rec
Forecast Change
4

Forecast Change
1
0.5

3
3-Quarte Change

2-YearChange

0
2
–0.5

8-Quarte Change
1

–1
1-YearChange
0

–1.5
–2
–1

0
Cur ent D via o f

2

1

–1
–2

0

–1

m Midpo nt f Ra ge

CurentYe

Bank of E gl d I ti F rec s

2

1

3

ar-Over-YPInflaetiroC

Gre n Bo k I flati F c s

Forecast Change
1.5

7-QuarteFocsChange
1.00
0.50

1
8-Quarte Change

0.00

0.5

4.5% “Target

–0.50

0

Sept mber
1990

–1.00
3-Quarte Change

–0.5

”

–1.50

May2003

–2.00
–1

–2.50

–1.5
2

1
CurentYear-Over-Yea

3

4

r RPIX Inflatio

the eight-quarter-ahead forecasts (depicted in the
lower left-hand panel) lie very close to the –45 degree
line, which is perhaps to be expected given the way
in which the Bank of England’s forecasts incorporate
recent policy actions. The Riksbank’s forecasts in
the upper right-hand panel are considerably more
spread out, in part reflecting the experience in the
mid-1990s when inflation was above its target and
expected to rise still further. For all three inflation
targeters, the three-quarter-ahead forecast inflation
changes generally lie some distance from the –45
degree line, suggesting that most deviations from the
inflation target are not expected to be fully reversed
at such a short horizon.
The Fed’s inflation forecasts, shown in the lower
right-hand panel, display an interesting pattern.
98

2.75% “Target

–3.00

J U LY / A U G U S T 2 0 0 4

2

4
3
CurentYe

”
5

6

ar-Over-YnflatioerI

Prior to December 1990, the long-horizon Green
Book forecasts tended to revert to a “target” of
approximately 4.5 percent, depicted by the higher
of the two –45 degree lines. (As noted above, this
“target” reflects the assumptions implicit in the
Green Book, rather than the intentions of the FOMC.)
But as inflation fell during the course of the recession, the implicit “target” in the Green Book seems
to have shifted down to roughly 2.75 percent.

Instrument Rules: Two Conventional
Specifications
This section presents estimates of conventional
instrument rules for the four central banks under
study. We begin with the simplest of the simple rules:

FEDERAL R ESERVE BANK OF ST. LOUIS

Kuttner

Table 1
Estimates of the Taylor Specification of the Instrument Rule
Dependent variable = policy interest rate, i
Coefficient on
Adjusted R2

LM test for
2nd order
autocorrelation

N

Intercept

Output gap

Inflation

New Zealand
1997:Q4–2003:Q2

23

1.14
(0.78)

0.25
(0.17)

–0.04
(0.23)

0.83***
(0.13)

0.64

12.6
0.001

Sweden
1994:Q1–2003:Q2

38

0.17
(0.17)

–0.63***
(0.13)

0.27***
(0.05)

0.99***
(0.04)

0.96

8.43
0.014

United Kingdom
1997:Q4–2003:Q2

23

–0.79
(0.61)

–0.43*
(0.22)

0.27
(0.28)

1.12***
(0.11)

0.90

3.25
0.197

United States
1987:Q1–1996:Q4

40

0.68***
(0.33)

0.32***
(0.08)

0.10
(0.14)

0.81***
(0.07)

0.95

7.06
0.029

Lagged i

NOTE: ***/**/* Indicate significance at the 1/5/10 percent levels, respectively. Estimation is by ordinary least squares. Numbers in
parentheses are standard errors. The LM test statistic for second-order autocorrelation is N times the R2 from a regression of the
residual onto all the regressors from the original regression and two lags of the residual. The p value from the χ2 distribution is reported
underneath the test statistic. For Sweden and the United Kingdom, the policy rate corresponds to the repo rate. For New Zealand, it
is the 90-day bank rate, and for the United States, it is the federal funds rate. For New Zealand and the United Kingdom, the inflation
rate used in the regression is the current-quarter “forecast” of four-quarter inflation. For Sweden, it is the four-quarter percentage
change in the CPI. For the United States, it is the four-quarter percentage change in the CPI, excluding food and energy. Proxies for
the output gap are constructed by accumulating the difference between the forecast growth rates of real GDP and an assumed rate
of potential growth. The data appendix contains additional details.

the classic Taylor (1993) rule, giving the policy rate
as a function of the current output gap and inflation
deviation,
(4)

it=(1 – ρ)i*+β xt,t+γ (πt – π*)+ρ it –1+et,

where xt,t is the current estimate of the current output gap, πt is the current four-quarter inflation rate,
π* is the target inflation rate, and i* is the steadystate nominal rate of interest (i.e., the inflation target
plus the equilibrium real rate of interest). The only
addition to the canonical Taylor specification is the
lagged interest rate term, whose conventional interpretation is in terms of interest rate “smoothing”—
the partial adjustment of the policy instrument
toward some underlying target rate. In this parameterization, the long-run response of the nominal
rate to inflation is γ /(1– ρ).
This is the point at which empirical research
on policy rules must confront the uncomfortable
fact that no central bank, except the RBNZ, consistently reports an estimate of the output gap, xt,t .27
27

The RBNZ reported quarterly output gap projections in its Monetary
Policy Statements from December 1997 through November 1999 and
again from December 2000 through March 2001. Annual averages
have been published consistently throughout the 1997-2003 period.

As a practical matter, this obviously complicates
any effort to model central bank behavior in terms
of a policy rule involving the output gap.
Rather than abandon the task at this point, we
proceed by constructing a proxy for the output gap.
Our approach is to back out an implicit estimate of
the output gap using the central banks’ projections
of real GDP growth. Two assumptions are needed
to make this work: (i) an estimate of the growth rate
of potential output and (ii) a terminal condition, i.e.,
the value of the output gap at the end of the forecast
horizon. For (i), we make an educated guess as to
the assumed growth rate of potential output, using
information gleaned from published central bank
sources. For (ii), we assume the output gap reverts
to zero at the end of the forecast horizon, unless
published information suggests otherwise. With
these assumptions, a real-time estimate of the output
gap can be constructed as the accumulated difference between the forecasts of real GDP growth and
the assumed potential growth rate. Further details
on the construction of the output gap proxy can be
found in the appendix.
The results, shown in Table 1, suggest a simple
rule like (4) is a poor description of policy for all
J U LY / A U G U S T 2 0 0 4

99

REVIEW

Kuttner

Table 2
Estimates of the Forward-Looking Instrument Rule
Dependent variable = policy interest rate, i
Coefficient on

LM test for
2nd order
autocorrelation

Current
output gap

Growth
forecast

Inflation
forecast

1.20
(0.61)

0.42**
(0.19)

0.50*
(0.26)

1.22**
(0.53)

0.75***
(0.10)

0.78

9.00
0.011

38

0.06
(0.41)

–0.18
(0.13)

0.31**
(0.12)

0.65***
(0.10)

0.77***
(0.05)

0.97

3.95
0.138

United Kingdom
1997:Q4–2003:Q2

23

–2.34
(1.39)

–0.06
(0.32)

0.53
(0.39)

0.09
(0.37)

1.19***
(0.13)

0.90

4.68
0.096

United States
1987:Q1–1996:Q4

40

–0.62
(0.52)

0.39***
(0.08)

0.20
(0.17)

0.50***
(0.15)

0.70***
(0.06)

0.96

4.61
0.100

N

Intercept

New Zealand
1997:Q4–2003:Q2

23

Sweden
1994:Q1–2003:Q2

Lagged i

Adjusted R2

NOTE: ***/**/* Indicate significance at the 1/5/10 percent levels, respectively. Estimation is by ordinary least squares. Numbers in
parentheses are standard errors. For New Zealand, Sweden, and the United Kingdom, the inflation forecast is the forecast change in
the target inflation rate over the subsequent four quarters, minus the inflation target (or the midpoint of the range, in the case of
New Zealand). For the United States, it is simply the forecast percentage change in the CPI over the subsequent four quarters, so the
regression intercept also includes any implicit inflation target. The growth forecast is for real GDP growth over the subsequent four
quarters, or in the case of New Zealand, the change in the output gap. See also notes to Table 1.

four central banks. For New Zealand, the coefficient
on inflation is effectively zero, and none of the
coefficients (except that on the lagged interest rate)
is statistically significant. The results are slightly
better for Sweden, which at least exhibits a statistically significant and positive response to inflation.
The coefficient on the output gap has the wrong
(negative) sign, however, and it is statistically significant. The results are also unsatisfactory for the United
Kingdom. Like Sweden, the output gap coefficient is
significant, but has the wrong sign. The inflation
coefficient has the right sign, but it is statistically
insignificant.28 In the United States, we find a significant, positive coefficient on the output gap, but the
coefficient on inflation is small and insignificant.
The estimated ρ parameter on the lagged interest
rate is large and highly significant in all cases, ranging from 0.81 for the United States to 1.12 for the
United Kingdom, suggesting an implausibly high
degree of interest rate smoothing. The R2 are high—
but only because of the explanatory power of the
lagged interest rate.
An alternative to Taylor’s specification, in the
28

One reason for the poor results could be the lack of any significant
variation in inflation or the output gap since 1997. Ironically, the
success of the Bank of England’s IT policy seems to have made it
more difficult to estimate a policy rule.

100

J U LY / A U G U S T 2 0 0 4

spirit of Clarida, Galí, and Gertler (2000), is to use
forecasts of inflation and the output gap, or alternatively the current output gap estimate and the
real GDP forecast,
(5)

it=(1 – ρ)i*+β1 xt,t+β2 ∆yt+k,t
+γ (πt+k,t – π*)+ρ it –1+et ,

where ∆yt+k,t and πt+k,t are the central bank’s forecasts of real GDP growth and inflation over the subsequent k quarters. This specification is attractive,
as it captures rational, forward-looking behavior on
the part of the central bank. In addition, by explicitly
including a forecast of inflation, the coefficients on
the output terms are readily interpreted in terms of
an output stabilization objective. By contrast, the
output gap may appear in the Taylor specification
as a predictor of inflation and may not imply a
weight on output stabilization per se.
Results from estimating (5) appear in Table 2,
with the horizon k set to four quarters. The results
are generally somewhat better than for the backwardlooking Taylor specification. The equation works very
well for New Zealand, in the sense that the coefficients on the current gap, the growth forecast, and
expected inflation are all significant and have the
expected signs. It also works reasonably well for
Sweden, where the parameter estimates on output

FEDERAL R ESERVE BANK OF ST. LOUIS

growth and year-ahead inflation are both highly
significant.29 The United Kingdom still yields poor
results, however. None of the coefficients is significant, although the coefficients on forecast GDP
growth and inflation at least have the correct signs.
For the United States, the output gap and the inflation forecast both have the correct signs and are
significant; in fact, the estimates look a lot like those
reported in Clarida, Galí, and Gertler (2000). The
Taylor principle seems to be satisfied for all but the
United Kingdom. Interestingly, the implied long-run
response of the interest rate to inflation is much
larger for New Zealand and Sweden (4.88 and 2.82,
respectively) than for the United States (1.34). But
this may reflect a difference in the nature of the
Green Book forecast which, unlike the inflation
targeters’, has not assumed a reversion of inflation
toward an unchanged target. The estimated coefficients on the lagged interest rates are still quite large,
ranging from 0.70 for the United States to 1.19 for
the United Kingdom.

Kuttner

for the superiority of formulating policy on the basis
of targeting rules. Omitted, serially correlated judgment terms could, therefore, lead to a spuriously
high degree of measured interest rate smoothing.
Our solution to this issue is a novel approach
to estimating an instrument rule that, in principle,
eliminates the serial correlation in the omitted
judgment term and allows for a clearer distinction
between interest rate smoothing and reaction to
this omitted variable. The innovation is to examine
the response of the policy rate to the “news” contained in the revisions in expectations embodied in
the central banks’ inflation and output forecasts.
This approach is derived quite simply by first
adding a lagged interest rate term to the instrument
rule (6),
(7)

it=(1 – ρ)i*+β xt+k,t+γ (πt+k,t – π*)
+φ zt+k,t+ρ it –1+et ,

and projecting it onto information available to the
central bank at time t –1,

Instrument Rules: A “Reaction to News”
Specification

(8)

One of the consistently unsatisfying results
from conventional instrument rule estimates, like
those presented above, is that the large estimates
of ρ imply an extremely high degree of interest rate
smoothing. Rudebusch (2002) argues that this reflects
the omission of highly serially correlated variables
from the instrument rule, rather than interest rate
smoothing per se.
In a similar vein, Svensson (2003) shows that
the optimal instrument rule for an inflation forecast
targeter has the form

Subtracting (8) from (7) gives

(6)

it=(1 – ρ)i*+β xt+k,t+γ (πt+k,t – π*)
+φ zt+k,t+et,

where zt+k,t is the central bank’s forecast of some
appropriate “judgmental” adjustment term, which
represents an omitted variable in the conventional
instrument rule specification.30 Indeed, as discussed
above in section 2, it is the presence of these unobserved judgment terms that leads Svensson to argue
29

Very similar results are reported in Berg, Jansson, and Vredin (2002),
who also find that the forecast of output growth is significant.

30

Svensson’s equation actually contains separate terms involving linear
combinations of one- and two-period-ahead forecasts of z, interpreted
as a vector of judgment terms. But for our purposes, a single z term
suffices.

it,t –1=(1 – ρ)i*+β xt+k,t –1+γ (πt+k,t –1 – π*)
+φ zt+k,t –1+ρ it –1+et,t –1.

(9) it – it,t –1=β (xt+k,t – xt+k,t –1) + γ (πt+k,t – πt+k,t –1)
+φ (zt+k,t – zt+k,t –1)+et – et,t –1.
Thus, the difference between the current policy
setting, it , and the expected policy setting as of the
previous period, it,t –1, depends entirely on the revisions in the k-period-ahead expectations of the output gap, inflation, and the unobserved z. Taking the
difference relative to the previous period’s expectation makes the interest rate smoothing term, ρit –1,
disappear. More importantly, if the central bank’s
forecasts are rational, then the unobserved revision
in the expectation of z should, by the law of iterated
projections, be unforecastable—and consequently
serially uncorrelated. Similarly, any residual serial
correlation in the et term will also disappear, leaving
a specification free from the usual econometric
problems that normally afflict estimated instrument
rules.31
Access to published central bank forecasts makes
it quite easy to implement this approach. Because
of the overlapping nature of these forecasts, the
31

While this specification resembles that of English, Nelson, and Sack
(2002), what distinguishes our approach is its emphasis on forecast
revisions.

J U LY / A U G U S T 2 0 0 4

101

REVIEW

Kuttner

Table 3
Volatility of Forecast Revisions and Interest Rate Changes
Standard deviation of
Four-quarter-ahead
real GDP forecast revision

Four-quarter-ahead
inflation forecast revision

Quarterly change
in policy rate

New Zealand, 1997:Q4–2003:Q2

0.58

0.30

0.78

Sweden, 1994:Q1–2003:Q2

0.34

0.32

0.48

United Kingdom, 1997:Q4–2003:Q2

0.35

0.31

0.40

United States, 1987:Q1–1996:Q4

0.41

0.28

0.53

NOTE: The forecast revisions represent the difference between the four-quarter-ahead forecast made in quarter t and the five-quarterahead forecast made in quarter t –1. The real GDP and inflation forecasts are both for four-quarter growth rates. For New Zealand,
the figure reported for the policy rate is the standard deviation of the policy rate forecast error, it – it,t –1. The standard deviation of
the raw change in the policy rate, ∆it , is 0.80.

revisions in expectations are calculated quite simply
as the difference between the current quarter’s kperiod-ahead forecast and the previous quarter’s
k+1-period-ahead forecast. This approach also
ameliorates the problem with the unobserved output gap. The output gap forecast can be written as
the difference between a forecast of real GDP and a
forecast for potential output—or, equivalently, the
difference between accumulated real GDP growth
and potential growth forecasts. If one is willing to
ignore quarterly revisions in potential growth forecasts on the grounds that these are likely to be small
relative to the forecast revisions in output growth,
then the accumulation of the revisions to real GDP
growth forecasts can be substituted for the output
gap forecast revisions.
The only remaining issue is what to use for it,t –1,
the previous period’s forecast of the current interest
rate. For the RBNZ, the solution is quite simple, as
the interest rate projections underlying the inflation
projections are published. In this case, it,t –1 is just
data. The solution is also straightforward for the
Riksbank and the Bank of England. If these banks’
stated policies of maintaining a constant interest rate
over the forecast horizon were accurate (and credible), then it,t –1 would simply equal it –1, leaving
(10) ∆it=β (xt+k,t – xt+k,t –1)+γ (πt+k,t – πt+k,t –1)
+φ (zt+k,t – zt+k,t –1)+et – et,t –1,
i.e., the change in the policy rate as a function of
“news” about inflation and the output gap. The
resemblance to Hall’s (1978) formulation of the
permanent income hypothesis is not coincidental.
102

J U LY / A U G U S T 2 0 0 4

Like a rational consumer, a central bank formulating
policy on the basis of efficient forecasts of its target
variables should satisfy an analogous set of orthogonality conditions. As a result, the central bank
should respond only to the news embodied in its
forecast revisions, and not to information that was
already known at time t –1.
For the reasons discussed above, however, the
assumption of a constant interest rate is not realistic,
even for those central banks nominally adhering to
a constant interest rate rule. It is nonetheless possible
to derive a specification similar to (10) for the case
of a time-varying (but unpublished) interest rate path.
To do so requires adding it,t –1 – it –1 to both sides of
(9), so that ∆it appears on the left-hand side; assuming xt+k –1,t –1 ≈ xt+k,t –1 (and likewise for π and z), the
it,t –1 – it –1 on the right-hand side becomes ρ ∆it –1 –
et –1+et,t –1. With this modification, equation (9)
becomes
(11)

∆it=β (xt+k,t – xt+k,t –1)+γ (πt+k,t – πt+k,t –1)
+φ (zt+k,t – zt+k,t –1)+ ρ ∆it –1+et – et,t –1,

the main difference from (10) being the reappearance
of the interest rate smoothing term, ρ ∆it –1.
Table 3 reports the standard deviation of the
forecast revisions and the change in the policy rate.
Inflation forecast revisions are very similar in
magnitude across the four central banks. Output
growth forecasts are somewhat more volatile for
New Zealand, which may explain the relatively high
degree of variability in that country’s short-term
interest rate. The United Kingdom’s interest rate, on
the other hand, is considerably less volatile, despite

FEDERAL R ESERVE BANK OF ST. LOUIS

Kuttner

Table 4
Estimates of the “Reaction to News” Specification of the Instrument Rule
Dependent variable = change in policy rate: ∆it , or it – it,t–1 for New Zealand
Coefficient on

New Zealand
1997:Q4–2003:Q2

Sweden
1994:Q1–2003:Q2

United Kingdom
1997:Q4–2003:Q2

United States
1987:Q1–1996:Q4

LM test for
2nd order
autocorrelation

∆it –1 or it –1 – it –1,t –2

Adjusted R2

1.45***
(0.46)

—

0.39

1.06
0.303

0.26
(0.22)

1.61***
(0.45)

0.33*
(0.17)

0.49

7.30
0.026

38

0.27
(0.22)

0.47**
(0.24)

—

0.09

13.00
0.002

38

0.37**
(0.16)

0.51***
(0.16)

0.56

10.00
0.007

23

0.39
(0.26)

0.73**
(0.35)

0.11

7.19
0.027

22

0.58**
(0.24)

0.52
(0.32)

0.32

2.22
0.330

37

0.02
(0.21)

0.55*
(0.31)

0.03

13.57
0.001

37

0.02
(0.19)

0.46
(0.27)

0.23

7.20
0.027

N

Output news

21

0.36
(0.22)

20

Inflation news

0.69***
(0.11)
—
0.47**
(0.19)
—
0.43***
(0.14)

NOTE: ***/**/* Indicate significance at the 1/5/10 percent levels, respectively. Estimation is by ordinary least squares. Numbers in
parentheses are standard errors. The “news” regressors are the central banks’ four-quarter-ahead forecast revisions: e.g., the time-t
forecast of four-quarter GDP growth four quarters ahead, minus the time-t–1 forecast of four-quarter GDP growth five quarters ahead.
See also notes to Tables 1 and 2.

having output and inflation forecast revisions comparable to those of Sweden.
The regression results appear in Table 4, with
and without the lagged interest rate. As in the conventional forward-looking reaction function estimates above, the horizon k is set to four quarters.
This specification works very well for New Zealand,
with a highly significant response to inflation news
and a positive (but not quite significant) response to
GDP news. In addition, the specification eliminates
the serial correlation in the residuals, with an LM
test statistic for second-order serial correlation of
only 1.06. (The lagged interest rate revision is only
marginally significant when it is included, although
the residuals appear more highly serially correlated
in this case.) The specification also works well for
Sweden, with a significant, positive response to both
output and inflation news; unlike New Zealand, the
coefficient on the lagged interest rate term remains
quite large, even in this specification. Moving to the
“reaction to news” specification greatly improves

the U.K. estimates: the coefficients on output and
inflation news have the right sign, the magnitudes
are plausible, and (depending on the specification)
the responses are statistically significant. The coefficient on the lagged interest rate is much smaller
(less than 0.5, rather than above 1.0), and there is
no evidence of serial correlation in the residuals.
The specification works least well for the United
States, where only the inflation coefficient is even
marginally significant and the United States R2 is low.
One successful aspect of the “reaction to news”
specification is the large reduction in the coefficient
on the lagged interest rate for the Bank of England,
the RBNZ, and the Fed. This result supports
Rudebusch’s (2002) contention that at least some
of the serial correlation in conventionally specified
instrument rules represents a response to an omitted
variable, rather than interest rate smoothing per se.
Indeed, for the United States the estimated lag coefficient of 0.43 is very close to the 0.4 Rudebusch
obtains from term structure estimates of funds rate
J U LY / A U G U S T 2 0 0 4

103

REVIEW

Kuttner

predictability, and it implies a much more plausible
level of interest rate smoothing. It is also not far from
the smoothing coefficient of 0.54 reported by
English, Nelson, and Sack (2002) for the forwardlooking specification with real-time data. The
Riksbank is an exception in this regard: The estimated lag coefficient remains a highly significant
0.7 even in the “reaction to news” specification,
perhaps suggesting a larger role for interest rate
smoothing in that case.32
Another attractive feature of this specification
is that it allows us to get a much clearer picture of
the degree of “judgment” in the setting of policy,
relative to a simple instrument rule. An assessment
of the role of judgment was difficult to make in the
conventional specifications, because it was hard to
distinguish between the omitted, serially correlated
z and interest rate smoothing. By contrast, the residual from equation (10) or (11) has a clearer interpretation in terms of revisions to the relevant judgment
factors.
The interpretation is particularly clean in the
case of New Zealand, where the non-constant
interest rate assumption and the availability of data
on it,t –1 mean the results are not muddled by the
implausibility of the constant interest rate assumption. In this case, the respectable but still relatively
low R2 (0.39 without the lagged interest rate, 0.49
with the lagged rate) means no more than one-half
of the variance in the interest rate can be traced
directly to revisions in the output and inflation forecasts. Thus, even the (arguably) most rule oriented
of all the IT central banks apparently still exercises
a great deal of judgment in setting policy.

What About Optimal Targeting Rules?
An alternative way to describe central banks’
behavior is in terms of a targeting rule, as opposed
to the instrument rules considered in the preceding
section. As stressed by Svensson (2003), this formulation has the advantage of being immune to the
inclusion of judgment terms in the central bank’s
forecast, and in the size of the policy action needed
to attain its target. While setting policy to achieve
an inflation target at a given horizon might be considered one form of an ad hoc targeting rule, the
question taken up in this section is whether inflation
targeters’ behavior can be characterized in terms of
simple optimal targeting rules.
32

The estimate of ρ shrinks if the sample is started in 1997, however,
suggesting that some of the apparent “smoothing” may be associated
with the dramatic decline in interest rates in 1996.

104

J U LY / A U G U S T 2 0 0 4

With a backward-looking inflation process in
which the inflation rate at time t+1 depends on the
current output gap, xt , the specific targeting rule
can be written as
(12)

πt+τ,t – π*=(λ/αx ) (xt+τ,t – xt+τ –1,t ),

i.e., as a linear relationship between the forecast
inflation gap in period t+τ and the forecast change
in the output gap between periods t+τ – 1 and
t+τ.33 (Note that here, the period corresponds to
the length of time it takes for a change in the output
gap to affect inflation, rather than a calendar quarter.)
The constant of proportionality is the ratio of λ, the
weight on output in the loss function, to αx , which
is effectively the slope of the aggregate supply relation (i.e., ∂πt+1/∂πt ).
The intuition behind this condition is straightforward and can be put in terms of marginal costs
and benefits. Take the case of τ=2,
(13)

πt+2,t – π*=(λ/αx ) (xt+2,t – xt+1,t ),

for example. Suppose inflation in period t+2 were
forecast to come in above its target, π*. With a quadratic loss function, the marginal benefit of reducing
inflation is simply equal to the gap between inflation
and its target. With a backward-looking inflation
process and a one-period lag in the response of
inflation, reducing inflation in period t+2 requires
running a negative output gap at time t+1, relative
to the future gap. In terms of output, the cost of
reducing inflation is proportional to 1/αx ; this
increases the loss function by an amount λ /αx. At the
optimum, the marginal benefit in terms of inflation
reduction is equated to the marginal cost of foregone
output.34
The bottom line is that under the assumption
of a backward-looking inflation process, optimal
monetary policy should induce a positive correlation
between the forecast inflation gap and the forecast
change in the output gap. This is not true if the
inflation process is assumed to be forward-looking,
however. In this case, the relevant targeting rule is
(14)

πt+τ,t – π*=–(λ/αx ) (xt+τ,t – xt+τ –1,t ),

and optimal policy induces a negative correlation
33

See Svensson (2003, equation 5.13). Note that this reflects the approximation that the discount factor, δ, is close to unity.

34

This explanation is admittedly somewhat heuristic. See Svensson (2003)
for a more precise (but arguably less intuitive) explanation in terms
of the relevant marginal rates of substitution and transformation.

FEDERAL R ESERVE BANK OF ST. LOUIS

between the forecast inflation gap and the forecast
change in the output gap.
Indeed, the only difference between this rule
and the one for the backward-looking inflation
process is the negative sign multiplying the change
in the output gap. Although it may seem odd that a
simple alteration to the assumed behavior of inflation
should change the sign of the relationship describing optimal monetary policy, the intuition is straightforward. In the forward-looking inflation model,
inflation in period t+2 responds contemporaneously
to the period-t expectation of the output gap in
period t+2. Thus, faced with greater-than-desired
inflation in period t+2, policy will tighten, inducing
a negative output gap (relative to the previous period)
in that period in order to bring inflation down.
This particular specific targeting rule corresponds
to one derived under pre-commitment, or a “timeless” perspective. Under discretion, the targeting
rule would look like
(15)

πt+τ,t – π*=–(λ/αx ) xt+τ,t ,

and not exhibit the “history dependence” that is a
hallmark of the pre-commitment solution. In this
case, optimal policy induces a negative correlation
between the forecast inflation gap and the forecast
output gap itself, rather than its change.
The lack of a clear implication for even the sign
of the relationship between inflation and output
gap forecasts means it is hard to put any of these
targeting rules to the test empirically—after all,
either a positive or a negative correlation could be
interpreted as evidence that the central bank was
obeying a specific targeting rule of some sort. By
contrast, the signs of the coefficients in the optimal
instrument rules do not depend on the assumed
nature of the inflation process, although the magnitudes do depend on this assumption. In any case,
any Phillips curve fit to the data would likely contain
both forward- and backward-looking terms, undoubtedly making the relevant instrument rules more
involved than those summarized above.
Further complicating matters is the question of
where to put the error term in a regression used to
estimate relationships like (12), (14), or (15) and
what that error term actually means. After all, these
targeting rules should hold exactly in terms of the
central banks’ forecasts, assuming those forecasts
embody the assumption that policy behaves optimally; any judgment should already be incorporated
into the central banks’ forecasts. (In the context of
estimating an instrument rule, the vice of judgment

Kuttner

terms becomes a virtue, in that it at least suggests a
plausible rationale for an error term in the regression.) One potential source of error is mismeasurement of the forecast output gap, which, as described
above, can generally only be inferred (imperfectly)
from central banks’ published GDP forecasts. Another potential source is that central banks behave
suboptimally—although in this case it is not clear
whether the error term from any less-than-optimal
policy should be thought of as orthogonal to the
inflation or the output gap forecasts (if either).
Our crude but effective solution to this normalization problem is to report the correlation coefficients, rather than the estimated slope coefficients
from regressions. The benefit of this approach is
that the correlation coefficient does not depend on
the normalization; the cost, of course, is that it says
nothing directly about the size of the parameter of
interest, λ/αx . But at least the correlation can say
something about the “closeness” of the empirical
relationship implied by the targeting rule; and, given
that even the sign of the relationship is up for grabs,
this seems like a reasonable compromise.
Table 5 reports the correlations relevant to the
three targeting rules discussed above. The first two
are those between the forecast inflation gap and the
forecast change in the output gap pertaining to the
backward-looking inflation process and the forwardlooking inflation process under pre-commitment,
for the four- and eight-quarter-ahead horizons. Also
reported are those between the forecast inflation
and output gaps pertaining to the forward-looking,
discretionary case.
In a few cases, there is a significant, negative
relationship between forecasts of the inflation gap
and forecasts of the change in the output gap, consistent with optimal monetary policy with a forwardlooking inflation process. For New Zealand, there is
a correlation of –0.46 at the four- to eight-quarter
horizon; for the United Kingdom, a negative correlation of a similar magnitude is observed at the
shorter zero- to four-quarter horizon. The correlations are relatively weak and insignificant in the
case of Sweden.
The correlation is a surprisingly strong –0.69
for the United States at the four- to eight-quarter
horizon. Interestingly, this is almost entirely the
result of the early 1990s’ disinflation, when the
Green Book contained forecasts for above-average
(but falling) inflation along with a widening output
gap—a period typically thought of more in terms of
serendipitous “opportunistic disinflation” than as
J U LY / A U G U S T 2 0 0 4

105

REVIEW

Kuttner

Table 5
Output-Inflation Correlations Corresponding to Targeting Rules
Correlations between
(πt+k,t – π*), (xt+k,t – xt+k–4,t )

(πt+k,t – π*), (xt+k,t)

k=4

k=8

k=4

New Zealand
1997:Q4–2003:Q2

–0.054
(0.81)

–0.455
(0.03)

0.415
(0.04)

Sweden
1994:Q1–2003:Q2

–0.181
(0.27)

–0.144
(0.42)

0.082
(0.62)

United Kingdom
1997:Q4–2003:Q2

–0.419
(0.05)

0.303
(0.16)

0.325
(0.12)

United States
1987:Q1–1996:Q4

–0.383
(0.02)

–0.689
(0.00)

–0.263
(0.21)

NOTE: Numbers in parentheses are p values for the hypothesis that the correlation is zero, assuming ρ̂ (N – 2)1/2/(1–ρ̂ 2 )1/2 follows a t
distribution with N – 2 degrees of freedom, where N is the number of observations. Because the output gap estimates often rely on
the assumption that the eight-quarter-ahead gap is zero, the correlation between the inflation and output gap at this horizon is not
reported.

an expression of forward-looking optimal monetary
policy. In interpreting this result, it is important to
keep in mind that the correlation is calculated assuming a constant mean, which implies an unchanged
inflation target. Consequently, the negative correlation may reflect the downward shift in the Fed’s
inflation target that coincided with the 1991-92
recession.
The results are less supportive of the forwardlooking, discretionary specification involving the
level of the output gap forecasts. The correlation is
insignificant for the United Kingdom, Sweden, and
the United States. Only for New Zealand is the correlation significant, but in this case it has the “wrong”
(i.e., positive) sign.
Overall, the results of this exercise lend lukewarm support for describing inflation targeters’
(and the Federal Reserve’s) monetary policy in terms
of a simple targeting rule, with an underlying forward-looking view of the inflation process. Still,
with estimated correlation coefficients generally in
the –0.3 to –0.5 range (corresponding to values for
the R2, 0.09 to 0.25), the goodness of fit is mediocre
at best and somewhat worse than that of typical
estimated instrument rules. It is also important to
bear in mind that uncovering a negative correlation
between inflation and output forecasts does not
necessarily imply that policymakers were behaving
optimally. Such a correlation could conceivably
arise even with policy governed by a suboptimal or
106

J U LY / A U G U S T 2 0 0 4

ad hoc rule (or none at all). Furthermore, the simple
targeting rules considered here will be misspecified
if the underlying inflation process contains both
backward- and forward-looking elements. Extending
the exercise to more realistic models, along the lines
of Giannoni and Woodford (2003c), is clearly
worthwhile—although it is also worth noting that,
as the targeting rule becomes more complex, it
becomes less useful as a means of communicating
the policy trade-offs to the public.

5. CONCLUSIONS
The overall goal of this paper has been to bridge
the gap between the literature describing the practice of IT and the literature on monetary policy rules.
In an effort to dispel some of the terminological
confusion that has contributed to this gap, section 2
reviewed some of the alternative definitions of the
term “policy rule,” while section 3 proposed various
ways of interpreting IT within the context of these
rules.
Ultimately, the question of how IT shapes
monetary policy is an empirical one, however. And
here, the empirical results of section 4 suggest a
number of important conclusions. The first is that
simple instrument rules do provide a reasonable
approximation to the conduct of monetary policy,
both by inflation targeters and by the Federal Reserve.
Second, instrument rules based on forecasts of inflation and output perform better than those based

FEDERAL R ESERVE BANK OF ST. LOUIS

only on current output and inflation. Third, while
inflation targeters tend to exhibit a somewhat larger
response to inflation forecasts than the Federal
Reserve, inflation targeters are not “inflation nutters.”
Although the results depend somewhat on the specification used, forecasts of the output gap or GDP
growth do seem to influence interest rate decisions,
even controlling for the inflation outlook.
But these conventional instrument rules also
leave a great deal unexplained (or at least “explained”
only by the lagged interest rate). This observation
suggests a fourth conclusion, namely, that central
banks exercise a great deal of judgment or discretion
relative to these rules. Estimates of an instrument
rule from a novel “reaction to news” specification
indicate that the Reserve Bank of New Zealand
probably comes closest to a pure inflation forecast
targeter; but even then, less than half of the variance
from the planned interest rate path can be traced
to forecast revisions.
A fifth conclusion is that simple optimal targeting rules of the kind advocated by Svensson (2003)
do not provide a particularly good description of the
conduct of monetary policy for any of the central
banks considered. Fitting optimal targeting rules to
the data, however, is complicated by the fact that
central banks generally do not report output gaps,
much less forecasts of those gaps; moreover, theory
alone provides no clear guide as to the correct sign
of the correlation implied by targeting rules.
At the level of estimated targeting or instrument
rules, it is hard to draw sharp distinctions between
the behavior of the three bona fide inflation targeters
studied and that of the Federal Reserve. But, in
general, the connection between changes in the
short-term interest rate and forecast revisions is
looser for the Fed than for the full-fledged inflation
targeters. A sixth conclusion that could be drawn
from the results, therefore, is that the FOMC uses
even more judgment than the (significant) amount
exercised by the inflation targeters.
While they may disappoint those who view IT
purely in terms of a policy rule, these conclusions
will come as no surprise to those familiar with
the practice of IT, as described, for example, by
Heikensten and Vredin (2002). At the same time,
the lack of a sharp, qualitative difference between
the Fed’s behavior and that of the inflation targeters
will probably do little to alter the priors of skeptics,
such as Ball and Sheridan (2003), who contend the
policy makes little practical difference.35
35

See, for example, the discussion in Posen (2002).

Kuttner

A hypothesis left unexplored in this paper is
that the real impact of IT is not so much on central
bank policymaking, per se, as it is on the impact of
those policy decisions upon inflation expectations
in the markets and the broader public; that is, that
talk does matter after all. This interpretation of IT
suggests that the response of expectations to economic news, or to policy itself, should be more
“anchored” for inflation targeters than for lesstransparent non-inflation targeters, as suggested in
Kuttner and Posen (1999, 2001) and Levin, Natalucci,
and Piger (2004). If so, then a better place to look
for effects of IT would be in the financial markets,
and particularly in the prices of assets that embody
inflation expectations. Examining financial markets’
response to policy—and, in particular, to “discretionary” policy actions with no apparent connection
to the central banks’ forecasts—may prove informative as to whether IT can anchor expectations as
its proponents claim.

REFERENCES
Agénor, Pierre-Richard and Masson, Paul R. “Credibility,
Reputation, and the Mexican Peso Crisis.” Journal of Money,
Credit, and Banking, February 1999, 31(1), pp. 70-84.
Archer, David. “Are the Policy Rules Proposed in the
Literature Good Enough for Practical Use?” Paper prepared
for the Norges Bank workshop, “The Role of Monetary
Policy in Inflation Targeting Regimes: Theory Meets
Practice.” Oslo, May 5-6, 2003.
Ball, Laurence and Sheridan, Niamh. “Does Inflation
Targeting Matter?” NBER Working Paper No. 9577,
National Bureau of Economic Research, 2003.
Batini, Nicoletta and Haldane, Andrew G. “ForwardLooking Rules for Monetary Policy,” in J.B. Taylor, ed.,
Monetary Policy Rules. Chicago: University of Chicago
Press, 1999, pp. 157-92.
Berg, Claes; Jansson, Per and Vredin, Anders. “How Useful
Are Simple Rules for Monetary Policy? The Swedish
Experience.” Unpublished manuscript, Sveriges Riksbank,
2002.
Bernanke, Ben S.; Laubach, Thomas; Mishkin, Frederic S.
and Posen, Adam S. Inflation Targeting: Lessons from the
International Experience. Princeton, NJ: Princeton
University Press, 1999.
Bofinger, Peter. “Inflation Targeting: Much Ado About
Nothing (New).” Paper prepared for the annual meeting

J U LY / A U G U S T 2 0 0 4

107

REVIEW

Kuttner

of the Ausschuss für Geldtheorie und Geldpolitik des
Vereins für Socialpolitik. Frankfurt, Germany, February 25,
2000.

Targeting by Inflation Targeting Central Banks. Geneva
Reports on the World Economy, Special Report 2. London:
Centre for Economic Policy Research, 2003.

Boivin, Jean. “Has U.S. Monetary Policy Changed? Evidence
from Drifting Coefficients and Real-Time Data.”
Unpublished manuscript, Columbia University, 2003.

Friedman, Benjamin M. and Kuttner, Kenneth N. “A Price
Target for Monetary Policy? Lessons from the Experience
with Money Growth Targets.” Brookings Papers on
Economic Activity, 1996, (1), pp. 77-125.

Carare, Alina and Stone, Mark. “Inflation Targeting Regimes,”
IMF Working Paper No. 03/9, International Monetary
Fund, 2003.

Friedman, Milton. Essays in Positive Economics. Chicago:
University of Chicago Press, 1953.

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.

Galí, Jordi. “Targeting Inflation in an Economy with Staggered
Price Setting,” in N. Loayza and R. Soto, eds., Inflation
Targeting: Design, Performance, Challenges. Santiago,
Chile: Central Bank of Chile, 2002.

Clarida, Richard; Galí, Jordi and Gertler, Mark. “Monetary
Policy Rules and Macroeconomic Stability: Evidence and
Some Theory.” Quarterly Journal of Economics, February
2000, 115(1), pp.148-80.

Gavin, William T. “Inflation Targeting: Why It Works and
How To Make It Work Better?” Business Economics, April
2004, 39(2), pp. 30-37.

Debelle, Guy. “The Australian Approach to Inflation
Targeting.” Unpublished manuscript, Massachusetts
Institute of Technology, 2003.
Drazen, Allan and Masson, Paul R. “Credibility of Policies
versus Credibility of Policymakers.” Quarterly Journal of
Economics, August 1994, 109(3), pp. 735-54.
Drew, Aaron and Plantier, L. Christopher. “Interest Rate
Smoothing in New Zealand and Other Dollar Bloc
Countries.” Discussion Paper 2000/10, Reserve Bank of
New Zealand, 2000.
Eggertsson, Gauti B. and Woodford, Michael. “The Zero
Bound on Interest Rates and Optimal Monetary Policy.”
Brookings Papers on Economic Activity, 2003, (1), pp.
139-211.
English, William B.; Nelson, William R. and Sack, Brian P.
“Interpreting the Significance of the Lagged Interest Rate
in Estimated Monetary Policy Rules.” Finance and
Economic Discussion Series No. 2002-24, Board of
Governors of the Federal Reserve System, 2002.
Faust, Jon and Svensson, Lars E.O. “Transparency and
Credibility: Monetary Policy with Unobservable Goals.”
International Economic Review, May 2001, 42(2), pp.
369-97.
Fracasso, Andrea; Genberg, Hans and Wyplosz, Charles.
How Do Central Banks Write? An Evaluation of Inflation
108

J U LY / A U G U S T 2 0 0 4

Geraats, Petra. “Central Bank Transparency.” Economic
Journal, November 2002, 112(483), pp. 532-65.
Giannoni, Marc P. and Woodford, Michael. “Optimal Interest
Rate Rules: I. General Theory.” NBER Working Paper No.
9419, National Bureau of Economic Research, 2003a.
Giannoni, Marc P. and Woodford, Michael. “Optimal Interest
Rate Rules: II. Applications.” NBER Working Paper No.
9420, National Bureau of Economic Research, 2003b.
Giannoni, Marc P. and Woodford, Michael. “Optimal Inflation
Targeting Rules.” NBER Working Paper No. 9939, National
Bureau of Economic Research, 2003c.
Goldfeld, Stephen. Comment on S. De Vay, “Modeling the
Banking Firm: A Survey.” Journal of Money, Credit, and
Banking, November 1984, 16(4), pp. 609-11.
Greenspan, Alan. Statement before the Subcommittee on
Domestic Monetary Policy Committee on Banking,
Finance and Urban Affairs, U.S. House of Representatives,
July 28, 1988.
Hall, Robert. “Stochastic Implications of the Life Cycle
Permanent Income Hypothesis.” Journal of Political
Economy, December 1978, 86(6), pp. 971-87.
Heikensten, Lars and Vredin, Anders. “The Art of Inflation
Targeting.” Sveriges Riksbank Economic Review, 2002,
(4), pp. 5-34.

FEDERAL R ESERVE BANK OF ST. LOUIS

Kuttner

Herrendorf, Berthold. “Inflation Targeting as a Way of
Precommitment.” Oxford Economic Papers, July 1998,
50(3), pp. 431-48.

McNees, Stephen K. “A Forward-Looking Monetary Policy
Reaction Function: Continuity and Change.” New England
Economic Review, November/December 1992, pp. 3-13.

Huang, Angela; Margaritis, Dimitri and Mayes, David.
“Monetary Policy Rules in Practice: Evidence from New
Zealand.” Discussion Paper No. 18/2001, Bank of Finland,
2001.

Mishkin, Frederic and Schmidt-Hebbel, Klaus. “A Decade
of Inflation Targeting in the World: What Do We Know
and What Do We Need To Know?” in N. Loayza and R.
Soto, eds., Inflation Targeting: Design, Performance,
Challenges. Santiago, Chile: Central Bank of Chile, 2002.

Jansson, Per and Vredin, Anders. “Forecast-Based Monetary
Policy: The Case of Sweden.” International Finance,
Winter 2003, 6(3), pp. 349-82.
King, Mervyn. “Changes in UK Monetary Policy: Rules and
Discretion in Practice.” Journal of Monetary Economics,
June 1997, 39(1), pp. 81-97.
Kuttner, Kenneth N. and Posen, Adam S. “Does Talk Matter
After All? Inflation Targeting and Central Bank Behavior.”
Federal Reserve Bank of New York Staff Report No. 88,
1999.
Kuttner, Kenneth N. and Posen, Adam S. “Inflation,
Monetary Transparency, and G3 Exchange Rate Volatility,”
in M. Balling, E.H. Hochreiter, and E. Hennessy, eds,
Adapting to Financial Globalisation: International Studies
in Monetary Banking. Volume 14. London: Routledge,
2001, pp. 229-58.
Leitemo, Kai.“Targeting Inflation by Constant-Interest-Rate
Forecasts.” Journal of Money, Credit, and Banking, August
2003, 35(4), pp. 609-26.
Leeper, Eric M. “An Inflation Reports Report.” Sveriges
Riksbank Economic Review, 2003, (3), pp. 94-118.
Levin, Andrew; Natalucci, Fabio and Jeremy, Piger. “The
Macroeconomic Effects of Inflation Targeting.” Federal
Reserve Bank of St. Louis Review, July/August 2004,
86(4), pp. 51-80.
McCallum, Bennett. “Inflation Targeting and the Liquidity
Trap,” in N. Loayza and R. Soto, eds., Inflation Targeting:
Design, Performance, Challenges. Santiago, Chile: Central
Bank of Chile, 2002.
McCaw, Sharon and Ranchhod, Satish. “The Reserve
Bank’s Forecasting Performance.” Reserve Bank of New
Zealand Bulletin, December 2002, 65(4), pp. 5-23.
McNees, Stephen K. “Modeling the Fed: A Forward-Looking
Monetary Policy Reaction Function.” New England
Economic Review, November/December 1986, pp. 3-8.

Nikolov, Kalin. “Monetary Policy Rules at the Bank of
England.” Paper prepared for the European Central Bank
workshop, “The Role of Policy Rules in the Conduct of
Monetary Policy.” Frankfurt, Germany, March 11-12, 2002.
Orphanides, Athanasios. “Monetary Policy Rules Based on
Real-Time Data.” American Economic Review, September
2001, 91(4), pp. 964-85.
Pagan, Adrian. “Report on Modelling and Forecasting at
the Bank of England,” 2003; www.bankofengland.co.ul/
pressrelease/2003/paganreport.pdf.
Posen, Adam S. Commentary on G. Chortareas, D. Stasavage,
and G. Sterne, “Does It Pay To Be Transparent?
International Evidence from Central Bank Forecasts.”
Federal Reserve Bank of Saint Louis Review, July/August
2002, 84(4) pp. 119-26.
Reifschneider, David L.; Stockton, David J. and Wilcox,
David. “Econometric Models and the Monetary Policy
Process.” Carnegie-Rochester Conference Series on Public
Policy, December 1997, 47, pp. 1-37.
Rogoff, Kenneth. “The Optimal Degree of Commitment to
an Intermediate Monetary Target.” Quarterly Journal of
Economics, November 1985, 100(4), pp. 1169-89.
Romer, Christina D. and Romer, David H. “Federal Reserve
Information and the Behavior of Interest Rates.” American
Economic Review, June 2000, 90(3), pp. 429-57.
Rudebusch, Glenn and Svensson, Lars E.O. “Policy Rules
for Inflation Targeting,” in J.B. Taylor, ed., Monetary
Policy Rules. Chicago: University of Chicago Press, 1999,
pp. 203-46.
Rudebusch, Glenn. “Term Structure Evidence on Interest
Rate Smoothing and Monetary Policy Inertia.” Journal of
Monetary Economics, September 2002, 49(6), pp. 1161-87.
Simons, Henry. “Rules versus Authorities in Monetary

J U LY / A U G U S T 2 0 0 4

109

REVIEW

Kuttner

Policy.” Journal of Political Economy, February 1936, 44(1),
pp. 1-30.

Carnegie-Rochester Conference Series on Public Policy,
December 1993, 39, pp. 195-214.

Sterne, Gabriel. “Inflation Targets in a Global Context,” in
N. Loayza and R. Soto, eds., Inflation Targeting: Design,
Performance, Challenges. Santiago, Chile: Central Bank of
Chile, 2002.

Taylor, John B. “Recent Developments in the Use of Monetary
Policy Rules,” in C. Joseph and A.H. Gunawan, eds.,
Monetary Policy and Inflation Targeting in Emerging
Economies. Jakarta: Bank Indonesia, 2000.

Stone, Mark R. “Inflation Targeting Lite.” IMF Working
Paper No. 03/12, International Monetary Fund, 2003.

Truman, Edwin M. Inflation Targeting in the World Economy.
Washington, DC: Institute for International Economics,
2003.

Svensson, Lars E.O. “Inflation Targeting as a Monetary
Policy Rule.” Journal of Monetary Economics, June 1999,
43(3), pp. 607-54.
Svensson, Lars E.O. “The First Year of the Eurosystem:
Inflation Targeting or Not?” American Economic Review,
May 2000, 90(2), pp. 95-99.
Svensson, Lars E.O. “What Is Wrong with Taylor Rules?
Using Judgment in Monetary Policy Through Targeting
Rules.” Journal of Economic Literature, June 2003, 41(2),
pp. 426-77.
Taylor, John B. “Discretion versus Policy Rules in Practice.”

Vredin, Anders. Discussion of S. Honkapohja and K. Mitra
“Problems in Inflation Targeting Based on Constant
Interest Rate Projections.” Prepared for the conference,
“Expectations, Learning and Monetary Policy.” Eltville,
Germany, August 30-31, 2003.
Willams, John C. “Simple Rules for Monetary Policy.”
Federal Reserve Bank of San Francisco Economic Review,
2003, pp. 1-12.
Woodford, Michael. Interest and Prices: Foundations of a
Theory of Monetary Policy. Princeton, NJ: Princeton
University Press, 2003.

Appendix

CONSTRUCTING REAL-TIME OUTPUT
GAP PROXIES
As described in section 4, the method used to
construct estimates of the output gap involves
accumulating the difference between the forecast
real GDP growth rates and some assumed potential
growth rate, imposing a terminal condition on the
output gap at the end of the forecast horizon. This
recursive technique is adapted to the specifics of
each central bank’s forecasts, as described here.

The United Kingdom
The Bank of England consistently reports fourquarter real GDP and inflation forecasts at a horizon
of one to eight quarters ahead. Because the individual
quarterly GDP forecasts are not reported, it is not
possible to back-out a quarterly output gap series.
But the four-quarter growth figures can be used to
construct measures of the current gap and the four110

J U LY / A U G U S T 2 0 0 4

quarter-ahead forecasts. The recursion can be
described as follows:
xt+8=0
xt+k=xt+k+4 – ( g – ∆4 yt+k+4) for k=4 and 0,
where xt+k is the k-quarter-ahead output gap, g is
the growth rate of potential output (assumed to be
a constant 2.4 percent per year), and ∆4 yt+k is the
forecast four-quarter real GDP growth rate (based
on the reported mode of the forecast distribution).

Sweden
The Sveriges Riksbank consistently reports a
current-year forecast for inflation and output and
forecasts for the subsequent two “out” years. Because
of this structure, the effective forecast horizon varies
between 8 and 11 quarters. The terminal condition
on the output gap is imposed only on the longest,
the 11-quarter “benchmark” forecast. Letting xs+k,q
represent the output gap in the qth quarter of year

FEDERAL R ESERVE BANK OF ST. LOUIS

s+k, the gap for the first quarter of each year is
constructed recursively in a manner similar to that
of the Bank of England:
xs+3,1=0
xs+k,1=xs+k+1,1 – ( g – ∆ ys+k ) for k=2 to 0,
where ∆ ys+k is real GDP growth in year s+k. Subsequent quarters’ output gap estimates are constructed
by accumulating over the relevant forecast horizon
the forecast revisions relative to the previous quarter.
Potential growth was assumed to be 2.6 percent
throughout, except in 2000 when it was 2.8 percent
and 2002 and 2003, when values of 2.5 and 2.4
percent were used, respectively. The end-of-forecasthorizon gap was assumed to be zero, except in 1999,
when the March Inflation Report referred to underutilized resources at the end of the forecast period
and in 2000 when the March Inflation Report referred
to capacity restrictions. For these years, the end-offorecast-period gaps are assumed to be –1 and 1
percent, respectively.

New Zealand
For 12 of the quarters in the sample, the Reserve
Bank of New Zealand reports quarterly estimates
of the output gap. For the remaining 12 quarters,

Kuttner

estimates were constructed using a method similar
to that used for the United Kingdom, based on
reported four-quarter real GDP growth rate forecasts.

The United States
The Federal Reserve is unique in that the Green
Book reports forecasts for quarterly GNP and/or GDP
growth, which allows the quarterly implied path of
the output gap to be extracted. However the forecast
horizon varies between four and nine quarters,
depending on the date at which the forecast was
made. As in Sweden, the terminal condition on the
output gap is imposed only for the long-horizon
(eight- or nine-quarter-ahead) “benchmark” forecasts, and the estimated gap is constructed as
xt+T=0
xt+k=xt+k+1 – (g – ∆ yt+k ) for k=T–1 to 0.
The accumulated revisions in the quarterly GDP
growth forecasts are used to update the gap estimates
between “benchmarks.” Forecast data for the first
through fourth quarters are taken from the Green
Books dated January or February, May, August, and
October or November. Prior to January 1992, output
growth is based on real GNP; after this, real GDP is
used.

J U LY / A U G U S T 2 0 0 4

111

REVIEW

112

J U LY / A U G U S T 2 0 0 4

Commentary
Monika Piazzesi

DISCUSSION

T

aylor rules, or modifications of Taylor rules
such as those proposed by Clarida, Galí,
and Gertler (1999), provide useful tools to
describe the behavior of central banks. A large literature has estimated these rules and has investigated
conditions under which it may be optimal for
central banks to use them. The question asked in
Kenneth Kuttner’s paper is whether these rules are
also useful in the context of inflation targeting.
To estimate these rules, the paper uses a new
approach. The idea is to measure expected inflation,
Etπt+k , and the output gap, Et xt+k , with data on the
central bank’s own projections (where k is the
horizon of the projection).1 Not all central banks
publish their own projection numbers. Those who
do publish them tend to be inflation-targeters. New
Zealand and the United Kingdom started publishing
them in 1997, while Sweden started doing it in 1994.
The estimations in the paper suggest that projection
data really help. While estimated policy rules based
on actual inflation and the output gap perform
poorly over the 1990s, estimated policy rules based
on projection data do a lot better: insignificant coefficient estimates become significant, “wrong signs”
turn around, coefficients on inflation get larger than
1, and residuals become less autocorrelated.
Some of the results are hard to interpret in terms
of what they mean for the behavior of these central
banks. For example, the coefficient on output is
significantly different from zero for Sweden and the
United Kingdom. With policy rules that are based
on actual inflation or some imprecise measure of
expected inflation, the coefficient on output may
just be due to the fact that output forecasts future
inflation. With policy rules based on central bank
1

Orphanides (2003) also takes this approach.

projections, this argument no longer applies; the
projections already contain all relevant conditioning
information for predicting inflation. Therefore, the
output coefficients seem to suggest that these central
banks are reacting to output. But what do we conclude from this?
I am excited about the idea of looking at central
bank projections and expect that we will see many
papers in the future that use these data to estimate
policy rules or to look at other issues. In what follows,
I will discuss two reasons to be excited that are not
mentioned in the paper (section 2). The first reason
is practical and has to do with the usual problems
of measuring inflation and the output gap. The
second reason is that projections may help us in
modeling learning by central banks. I will also mention other issues that could be explored with the
data (section 3). At the end of my discussion, I will
bring up some disadvantages associated with central
bank projections (section 4). These include potential
incentives of central banks to manipulate their own
projection numbers. So far, the main drawback is
that the data sample is short. But the good thing
about samples is that they are like trees—they grow.

ADVANTAGES OF PROJECTION DATA
Practical Issues
There are several practical issues associated
with estimating forward-looking policy rules. It
starts with simple measurement problems. What is
the right measure of inflation? John Taylor has used
the consumer price index and the gross national
product deflator in his papers. At this conference,
Andrew Levin, Fabio Natalucci, and Jeremy Piger
have focused on core inflation, while Laurence
Meyer has advocated the core personal consumption expenditure (PCE) deflator. What is the right
measure of the output gap? At this conference, Lars

Monika Piazzesi is an assistant professor of finance at the University of Chicago and a faculty research fellow at the National Bureau of Economic
Research.
Federal Reserve Bank of St. Louis Review, July/August 2004, 86(4), pp. 113-15.
© 2004, The Federal Reserve Bank of St. Louis.

J U LY / A U G U S T 2 0 0 4

113

REVIEW

Piazzesi

Svensson has recommended the Kalman filter to
compute the output trend as opposed to the HodrickPrescott filter. But even apart from these problems,
it is not clear how we should be computing expected
inflation, Etπt+k , and the output gap, Et xt+k . To
compute these conditional expected values, we need
to pick conditioning variables and, more generally,
the dynamics of inflation, π, and the output gap, x.
Projection data offer an easy way out of all these
practical issues. If a central bank publishes PCE
projections, it must be because the board members
closely follow the evolution of this particular price
index. Similarly, the output-gap projections are based
on some kind of trend calculation, and we do not
need to find out how this was done exactly. The projections already condition on the relevant variables,
and there is no model to chose. If a central bank
publishes projections Etπt+k and Et xt+k for different
horizons, k, the only question is what horizon k to
pick to estimate the policy rule. But picking k seems
easy compared with the host of other problems
that we run into otherwise.

Learning by Central Banks
Tom Sargent and others have argued that learning by central banks is important to understand their
behavior. In a model with learning, the current belief
of the central bank is a state variable. If the data we
observe are generated from such a model, we are
bound to find parameter instability in policy rules
that are written in terms of current inflation, πt,
and the output gap, xt. Now projection data may
just be the right measure of current beliefs. If this
is right, we may be able to estimate policy rules
that are stable functions of the projection data.

MORE THINGS TO DO WITH
PROJECTION DATA
Model Behind the Projections
It would be interesting to know what a model
of the economy that gives rise to these projections
would look like. In particular, it would be interesting
to see whether learning is an important feature of
such a model. To set up such a model, answers to
the following two questions would be useful. First,
what are the empirical properties of the projection
data? The paper plots the data in Figures 1 and 2.
The paper also computes the variances of changes
in the projections in Table 3. But it would be useful
to know more about the data: Are these projections
114

J U LY / A U G U S T 2 0 0 4

unbiased? Are the projection errors autocorrelated?
Can they be forecasted with lagged macroeconomic
variables? How do the projections compare with
forecasts from estimated autoregressive processes?
How do they compare with those from estimated
vector autoregressions (VARs)?
Second, how do policy rules based on projection
data and on VAR forecasts compare? Clarida, Galí,
and Gertler (1999) compare their forward-looking
policy rule with rules based on πt and xt. The paper
here compares policy rules based on projection data
with rules based on πt and xt. Now it would be interesting to know how the policy rules here compare
with those estimated by Clarida, Galí, and Gertler.

Financial Data and Private Information
An alternative way to measure expectations is
to use financial data. I have looked at this issue in
the context of a model of the term structure of
interest rates, in which the Federal Reserve targets
the short rate (Piazzesi, forthcoming). According to
the estimated policy rule from the model, the Fed
reacts to information contained in the term structure,
which is available right before the Federal Open
Market Committee (FOMC) meeting. I document
that the rule from the model performs better than
Taylor-type rules, at least as a description of Fed
behavior.
One reason to estimate policy rules based on
yield data is that yields may be a good proxy for the
conditioning information available to the central
bank at the time of the policy decision. Financial
data, however, only reflect public information. If
private information of the central bank is important
for these policy decisions, projection data may be
preferable. To see whether private information
matters, one could compare rules based on these
two types of data.
Another interesting question would be to analyze
the yield-curve implications of a model that is able
to explain the projection data. If learning is part of
the story, it would be exciting to see how it shows
up in yields.

DISADVANTAGES
A disadvantage of projection data is that central
banks have started to publish them only recently.
The evidence presented in this paper is thus only
based on a short sample from the 1990s. But the
sample is growing, so that we will have more observations soon.

FEDERAL R ESERVE BANK OF ST. LOUIS

Piazzesi

Moreover, projection numbers may not be the
numbers that ultimately influence policy decisions.
For example, the Fed’s staff presents Green Book
forecasts to the FOMC. But, of course, the FOMC
has its own views about future economic developments, and its policy decisions are based on these
views.
Finally, central banks may have incentives to
distort their projection numbers. Such incentives
may be particularly strong for inflation-targeting
central banks, whose projection numbers are closely
watched by the public. For these banks, inflation
projections play similar roles to earnings projections
by private firms. I do not know whether these incentive problems are severe, but it is certainly something to keep in mind when we interpret the results
obtained with these data.

REFERENCES
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.
Orphanides, Athanasios. “Historical Monetary Policy
Analysis and the Taylor Rule.” Journal of Monetary
Economics, July 2003, 50(5), pp. 983-1022.
Piazzesi, Monika. “Bond Yields and the Federal Reserve.”
Journal of Political Economy (forthcoming).

J U LY / A U G U S T 2 0 0 4

115

REVIEW

116

J U LY / A U G U S T 2 0 0 4

Is Inflation Targeting Best-Practice Monetary
Policy?
Jon Faust and Dale W. Henderson

1. INTRODUCTION

T

he core requirements of inflation targeting
are an explicit long-run inflation goal and a
strong commitment to transparency. The
framework built around these requirements has
much to recommend it. Inflation and output performance in economies using the inflation-targeting
framework (ITF) has been good by historical standards, and both governments and central banks
claim to be pleased with the framework. Advocates
and practitioners of the ITF have been leaders in
shaping and exploiting the new consensus that
central bank transparency can make policy more
effective. Not only are ITF central banks among
the most transparent in the world, they have experimented aggressively with ways to make communication with the public more effective. In the process,
they have pioneered the use of various tools, such
as fan charts, that make conveying essential, but
difficult, concepts practical.
Economic performance in some non-ITF economies, such as the United States, has also been good
in recent years. However, several ITFers (Mishkin,
1999, and Bernanke et al., 1999) argue that this
outcome has resulted in spite of the policymaking
frameworks in those countries. Thus, they argue
that the United States and others should “fix the
roof while the sun is shining.”
The recent hurricane in Washington has
reminded many of us of the wisdom of this reasoning. It has also reminded us that even attractive new
roofs—roofs that have weathered a few spring rains—
might bear inspection. In that spirit, we examine
whether the ITF constitutes best-practice monetary
policy. We use the standard of some mythical bestpractice policy to emphasize that we are not simply

ranking the ITF relative to some set of ad hoc alternatives, such as the current practices at central banks
around the world.
Of course, the ITF community is now large and
varied. We focus mainly on the industrialized countries. One might say that we consider the role of
the ITF in a low-inflation steady-state and do not
address the important question of how it might help
in reaching such a steady-state. In much of the paper
we seek to highlight some generic issues. In doing
so, we do not mean to suggest that there are not
important differences among the practices of ITF
central banks: Our points will apply in varying
degrees to ITF and other central banks.1
Our main message can be summarized succinctly. Common wisdom and conventional models
suggest that best-practice policy can be summarized
in terms of two goals: First, get mean inflation right;
second, get the variance of inflation right. The ITF
is of great help in achieving the first goal; whether
it helps in achieving the second is more problematic.
The argument goes as follows. Everyone now agrees
that mean inflation should be modest. The ITF may
be seen as a constructive attempt to cement the
current consensus on this point. Unfortunately,
agreement regarding the mean inflation rate has
few practical implications at any finite horizon. Many
of the most contentious debates over the conduct
of policy in the postwar era are not about the mean
but about the variance of inflation. That is, under
what conditions should the central bank allow or
promote movements of inflation around the mean
in order to promote other goals such as real and
1

There are excellent taxonomies of the approaches used by different
ITF central banks. See, e.g., Bernanke et al. (1999), Debelle (2003), and
Truman (2003).

Jon Faust is assistant director and Dale W. Henderson is senior advisor in the division of international finance at the Board of Governors of the
Federal Reserve System. The authors thank, without implicating, Michael Bordo, Matthew Canzoneri, Jinill Kim, Eric Leeper, Athanasios Orphanides,
David Small, Lars Svensson, Peter Tinsley, Edwin Truman, Anders Vredin, and Alexander Wolman. They have benefited greatly from the comments
of Benjamin Friedman, the discussant for this paper at the conference. The views in this paper are solely the responsibility of the authors and should
not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System or any other person associated with the Federal
Reserve System.
Federal Reserve Bank of St. Louis Review, July/August 2004, 86(4), pp. 117-43.
© 2004, The Federal Reserve Bank of St. Louis.

J U LY / A U G U S T 2 0 0 4

117

REVIEW

Faust and Henderson

financial stability? We argue that the ITF does not
constitute best-practice in resolving this question.
This claim is not new; it has also been made by both
critics and supporters of the ITF, including Kohn
(forthcoming), Benjamin Friedman (2003), and
Svensson (1999).
Of course, the ITF emerged near the end of a
period of high inflation in the industrialized economies, when the most important challenge for policy
was getting the mean right.2 As a period of generally
low and stable inflation has emerged, more attention
has been focused on the potential role of other goals
in policy, and greater emphasis both in the academic
literature and in ITF practice has been placed on the
role of other goals. Our main critical point is not
simply that there is more left to do. Rather, we argue
that various features of the ITF—for example, the
way the preeminence of the inflation goal is stated—
obscure rather than facilitate the communication
of best-practice policy.
Although we talk a great deal about the second
goal of policy, getting the variance right, we emphasize that getting the mean right may be the goal of
greatest importance. Arguably, the largest mistakes
in the postwar era have been associated with failures
to achieve this goal. Further, as a profession we are
more certain about our advice regarding the mean
of inflation and more confident that central banks
can get it right. At a minimum, the best-practice
policy framework should stress the goal that we are
more clear about and that we are more confident
central banks can achieve.
Finally, we raise many issues that are in principle
subject to empirical investigation and, therefore,
may one day be resolved. We focus on issues not
currently amenable to clear empirical resolution.
We do raise some questions for future empirical
assessment.
In the next two sections we characterize the ITF
in more detail and discuss some claims about the
economy that underlie the approach. Our characterization, and our thinking more generally, rely
heavily on such seminal work as Bernanke and
Mishkin (1997) and Bernanke et al. (1999) and several contributions by Svensson including Svensson
(1997a). The following three sections deal at a fairly
2

Truman (2003) emphasizes the important distinction between actual
or potential inflation targeters that have essentially achieved their
desired mean inflation rates and those that have not. He provides a
thorough analysis of the experience with inflation targeting and argues
that mutual understanding among the major central banks would be
significantly increased if they all adopted inflation targeting.

118

J U LY / A U G U S T 2 0 0 4

abstract level with macroeconomic and politicaleconomy aspects of the ITF. In the final two sections,
we return to reality, discussing some complicating
factors missing from the earlier analysis and then
making some constructive suggestions.3

2. WHAT IS THE INFLATION-TARGETING
FRAMEWORK?
2.1 Central Elements
One summary view of the ITF is provided by
Bernanke et al. (1999, hereafter BLMP):
Inflation targeting is a framework for monetary policy characterized by the public
announcement of official quantitative targets
(or ranges) for the inflation rate over one or
more horizons, and by explicit acknowledgment that low and stable inflation is monetary policy’s primary long-term goal. Among
other important features of inflation targeting are vigorous efforts to communicate with
the public about the plans and objectives of
the monetary authorities, and, in many cases,
mechanisms that strengthen the central
bank’s accountability for attaining those
objectives. (p. 4, italics added)
We have italicized what we believe to be central
elements. At the most general level, these elements
can be summarized as follows: Set an explicit, longrun inflation goal, give that goal a certain preeminence, and communicate vigorously about the
conduct of policy relative to that goal. What may not
be clear from the passage above is that ITF advocates
also recommend taking into account goals other than
inflation—for example, real and financial stability—
and call for thorough communication about these
goals. All major ITF banks have such goals.
We will regularly refer to two core requirements
of the ITF: (i) set a long-run inflation goal and (ii)
strive vigorously for transparency regarding all goals
and aspects of policy.
These requirements might be viewed as
innocuous. A long-run inflation target need have
few finite-horizon implications. As Keynes (1923,
p. 80) famously put it in discussing earlier monetary
reforms,
But this long run is a misleading guide to current affairs. In the long run we are all dead.
3

Gavin (2004) also makes useful suggestions for improving the ITF
framework.

FEDERAL R ESERVE BANK OF ST. LOUIS

The remaining core requirement is transparency.
While how best to achieve transparency is a difficult
question, that transparency is important seems
uncontroversial; rather, it seems to be the new
orthodoxy.
Based on these core requirements alone, it would
be difficult to understand why the ITF has generated
such strong sentiment for and against.
Going beyond the core requirements, advocates
portray the full-blown ITF as a framework of “constrained discretion” and claim that it has great advantages over “purely discretionary” policy as practiced,
say, at the Federal Reserve. In this paper, we attempt
to identify which requirements of the ITF, beyond
the core, give rise to such claims and then discuss
whether these features should unambiguously be
included among the requirements for best-practice.

2.2 Rules Versus Discretion
One definitional question that arises immediately
is whether the ITF is best viewed as a “rule” or the
exercise of “discretion.” Proponents such as BLMP
and Svensson appear to differ on this matter, so it is
important for us to make our view clear at the outset.
BLMP contend that the ITF is “not a rule in the
classical sense” (p. 22); that is, it is not a forcing rule—
a constraint on behavior that cannot be circumvented.
They note that, “an inflation-targeting framework
will not directly prevent counterproductive attempts
of a central bank to apply short-run stimulus. In
this respect, inflation targeting is inferior to an ironclad rule” (p. 24).
In contrast, Svensson (1999) argues that the
ITF is a rule. We think that there is no contradiction
here: Svensson is simply using a different definition.
He mentions the standard definition, but provides
an alternative under which advice to the central
bank about how to use its discretion is a rule.4 For
purposes of this discussion, the important point is
that a Svensson-type rule does not represent even an
approximation to a forcing rule. Any consequences
of ignoring Svensson-type rules are contingent on
the reaction of the public. Thus, in our view the ITF
is a typical example of discretion in the classical
sense, and we treat it as such in what follows.
4

In simple models, there is a multitude of formally equivalent ways of
recommending that the central bank optimize; an example is “satisfy
your first-order conditions.” Svensson (1999) examines several interesting exact and approximate ways to recommend optimization in the
linear-quadratic framework.

Faust and Henderson

2.3 What the “Discretion” View Implies
for Analyzing the ITF
Much standard policy advice comes in the form
of concrete suggestions about how to pursue the
goals of policy. Such advice often is either codified
in the form of a reaction function or can reasonably
be so codified for the purposes of study. In these
cases, one can take some set of interesting macro
models and run horse races among the implied
reaction functions. One can also solve for the optimal
rule and measure the inefficiency of the proposed
rules relative to the optimum. Based on the results
from several models, one can make statements about
robustness and so on.
Many critics of the ITF have adopted a rules
interpretation, but BLMP argue that these exercises
completely miss the point (p. 21).5 BLMP argue along
the lines we have been discussing: The ITF does not
place any constraints on the central bank that force
it to deviate from the social optimum. Thus, it is not
sensible to compare ITF outcomes to the optimum—
the ITF can attain the optimum.6
It may seem like cheating to propose a framework and then simply to stipulate that it delivers
optimal policy in model-based horse races. What
role does this leave for policy research? We can distinguish two strands of the policy literature, which
might be called institutional design and day-to-day
implementation. If a forcing rule for central bank
behavior is feasible and optimal, these two collapse
to one: Implement the forcing rule. If judgment must
be exercised by the policymaker, these two interact
but can be clearly distinguished. Research on institutional design seeks to define the optimal framework within which judgment will be exercised.
Research on day-to-day implementation seeks to
generate insights that will help policymakers exercise
their judgment. The horse-race exercises may shed
little direct light on the question of institutional
design, but may be of great value in informing the
ITF policymaking board regarding how best to use
its discretion.
5

Examples are Friedman and Kuttner (1996), Jensen (2002), and Kim
and Henderson (2002).

6

This statement is correct, so long as the social optimum is consistent
with hitting a long-run inflation objective, as it is in virtually all horserace models. There are many possible optima given by various discretion and commitment solutions. Exactly which one is reached depends
on considerations (communication policy, etc.) that are generally left
out of horse-race models. As we will explain, it is these considerations
that are supposed to ensure that the ITF yields the right optimum.

J U LY / A U G U S T 2 0 0 4

119

REVIEW

Faust and Henderson

3. CONVENTIONAL MACROECONOMICS
AND THE ITF
The reasoning supporting the ITF rests on several
features of common ground in the macroeconomics
profession and the policy community. Ultimately,
some confusion about the ITF seems to arise because
there are ITFers on both sides of a familiar dispute.
In this section, we survey both the common ground
and the disputed ground.

3.1 Macro Common Ground
3.1.1 Best-Practice Monetary Policy Would
Deliver Low, Stable Inflation. Inflation should be

stationary, perhaps with infrequent, small mean
shifts. Its mean should be low, certainly above zero
and below 5 percent. Its variability should be sufficiently small that annual inflation is within 1 or 2
percentage points of the mean most of the time.
Many technical issues may arise in giving this claim
greater precision, but some version of it is now
nearly universally accepted. This claim rests on
three more basic claims.
First, there is no long-run Phillips-curve trade-off
of the traditional variety. Failure to recognize this
“fact” was surely behind some mistakes of the past,
and Friedman (1968) and Phelps (1968) deserve
great credit for pointing this out.7
Second, marginal positive expected inflation
above some low rate is welfare-reducing. The profession can now list a multitude of channels through
which expected inflation can affect welfare by reducing growth or in other ways. Widespread acceptance
of the second claim rests neither on an appeal to a
particular channel nor on incontrovertible econometric estimates of the costs of inflation. Nonetheless, in this paper we take for granted the common
position that, above fairly low levels, raising mean
inflation is bad.
Third, marginal decreases in expected inflation
below some low, positive level are welfare-reducing.
The recent experience of Japan and the return of
low inflation in many countries has led economists
to explore a number of channels through which
deflation could be harmful. We will not focus much
7

This first claim is sometimes stated as the claim that expected inflation
does not affect real variables in the long run. When stated this way,
the myriad ways that expected inflation leads to welfare losses are
presented as exceptions. Recently, considerable attention has been
focused on the possibility that expected inflation might affect output
because wages and prices are set in staggered contracts that are not
fully indexed. See, for example, Wolman (2001).

120

J U LY / A U G U S T 2 0 0 4

on deflation, but it is important to emphasize that
inflation costs are two-sided.
3.1.2 There Is a Conventional Short-Run
Phillips Curve Trade-off. There may be many
short-run trade-offs, but for concreteness we focus
on the one that is most familiar to and relevant
for many economists: the short-run Phillips-curve
relation. We will take it as given that real activity
is not always at the most efficient level and define
the output gap as actual output minus efficient output. Policy actions that increase the gap are associated with a rise in inflation relative to steady-state
inflation. As the next point makes clear, we do not
mean to imply that best-practice policy can successfully exploit this trade-off.
3.1.3 The Economy Is Complicated. Economic complications may include policy lags, changing relationships, the potential for self-fulfilling
equilibria, and other nonlinearities. Finally, describing the state of the economy may require a highdimensional state variable.
For a brief period, some leading members of
the profession believed we were reaching the point
where our understanding of economic dynamics
allowed considerable range for beneficial intervention in business cycle dynamics. Friedman argued
against this view, claiming that policy acted with
long, variable, and unpredictable lags.8 Building on
Friedman, Lucas famously added the Lucas critique
to the list of complications.9 The general view that
the economy is very complex is now widely accepted
by academics and policymakers, for example,
Greenspan (2003).
3.1.4 Due to Political-Economy Problems,
Institutional Design Matters. There are various
time-consistency, game theoretic, and institutional
problems that might cause the government to set
policy away from the social optimum.10 We lump all
these under the title of political-economy problems.
The profession still disagrees about the importance
of, for example, time-consistency problems over
the postwar era. But even those who feel that time
consistency was not a big issue probably agree that
institutional design is important. Some version of
the claim that governments may have an incentive
8

See Friedman (1948, 1959).

9

See Lucas (1976).

10

Kydland and Prescott (1977) and Barro and Gordon (1983) initiated a
large literature on time consistency. There are many good treatments
of the political-economy issues, e.g., Alesina and Rosenthal (1995) and
Persson and Tabellini (2002).

FEDERAL R ESERVE BANK OF ST. LOUIS

to exploit inflation surprises has been a common
belief since the first time a government debased
its currency.

3.2 Long-Disputed Ground
Whether monetary policy can beneficially
exploit the short-run trade-off between inflation
and the output gap is a long-standing dispute that
is still at the center of monetary policy discussions.
3.2.1 No Exploitable Trade-offs. Friedman,
and later Lucas, forcefully argue that given the
inherent complexity of the economy and our regrettably limited knowledge of it, the ambitions of
monetary policy should be limited to achieving
nominal stability. Their arguments are based on the
view that monetary policy has strong short-run real
effects but that there is no way monetary policy
can beneficially exploit them. They both suggest
that a cautious response in the form of a k-percent
rule for money growth is the best way to achieve
nominal stability.
Some ITFers hold the no exploitable trade-offs
view (NET) and might well be called neoFriedmanites. The NETers follow Friedman and Lucas
in asserting that any trade-offs that exist cannot be
successfully exploited, so that best-practice can only
hope to achieve nominal stability. However, they
argue that inflation stabilization is the best way to
achieve nominal stability.11 They are at least as pessimistic as Lucas and Friedman about complications
in the economy. Nevertheless, they argue that achieving inflation stability requires judgment and looking
at a wide range of information variables (a “look at
everything” strategy) and may require frequent
changes in the instruments of policy.
Orphanides (2003b) provides support for the
NET view and discusses its roots in Friedman. Ernst
Welteke (2003), president of the Bundesbank and
member of the governing council of the European
Central Bank, has also clearly stated the NET view.
Mishkin (2002) and BLMP both state the case for
the NET view clearly, but, as we will explain, they
belong in another camp.
What we will call the singular economy view is
an alternative route to the NET position. In this view
the economy is stochastically singular in a helpful
way: Stabilizing inflation automatically achieves any
other goals of policy. For example, Rotemberg and
11

One might also imagine a neo-Friedmanite view centered on other
notions of nominal stability, such as nominal income stability. Given
the topic of the paper, we do not develop this idea.

Faust and Henderson

Woodford (1997), King and Wolman (1999), and
Goodfriend and King (2001) produce simple models
that exhibit a happy coincidence of the goals of
stabilizing inflation and the output gap.12
Thus, either a vexing complexity or a fortuitous
simplicity of the economy can get one to the view
that there are no exploitable trade-offs.
3.2.2 Limited Exploitable Trade-offs. Finetuning the real and nominal economy is overly
ambitious. In the limited exploitable trade-offs view
(LET), there is some beneficially exploitable shortrun trade-off between real activity and inflation,
and best-practice policy exploits it. LETers, like
NETers, contend that best-practice cannot be implemented using a rigid rule or by following a formal
model. Studying optimal policy in formal models
serves mainly to inform our collective wisdom, and
this wisdom should be applied deftly in practice.
Many, if not most, advocates of the ITF are LETers.
For example, Svensson and Woodford belong in the
LET camp: In numerous papers, such as Svensson
(1997a) and Svensson and Woodford (forthcoming),
they describe the ITF as involving optimal exploitation of the short-run trade off. BLMP also belong in
this camp. Despite the claim of President Welteke
of the Bundesbank cited above, we suspect that most
central banks are in the LET camp, but this empirical claim need not detain us.
3.2.3 Comments. In both views, no rigid rule
is appropriate, and policy must be based on a review
of a wide variety of information. Neither camp takes
an a priori stance on whether best-practice policy
is “activist” in the sense of requiring frequent
adjustment of instruments. Only in the LET camp
is policy “activist” in the sense of attempting to
manage the business cycle.
Because NETers argue that policy should only
aim for inflation stability, they are open to the criticism that they are inflation nutters—which we take
to mean that inflation is the only thing in the loss
function. This criticism is misplaced: NETers are not
nutters. The NETer argues from standard preferences
that achieving nominal stability is the best we can
hope for.
12

In these models, only the price of the single composite good is affected
by staggered contracts. There are at least two standard modifications
that imply trade-offs. The first is adding a cost shock to the price-setting
equation for the single good as in, for example, Kiley (1998), McCallum
and Nelson (1999), and Clarida, Galí, and Gertler (1999). This modification is incorporated into the simple model we will present here. The
second modification is to assume that the wage of composite labor
(or the price of a second composite good) is affected by staggered
contracts, as in, for example, Erceg, Henderson, and Levin (2000).

J U LY / A U G U S T 2 0 0 4

121

Faust and Henderson

While the NET and LET views are distinct, in
practice it is sometimes difficult to tell which view
various parties take. Several problems emerge. First,
some LETers believe that the degree of exploitability
is quite low; thus, the views need not be that far apart.
Second, virtually everyone agrees that demand
shocks push us toward the singular economy perspective. That is, in many standard models, demand
shocks temporarily increase the output gap and raise
inflation. Thus, smoothing inflation and the gap
suggest roughly the same response. While limiting
attention to demand shocks does not lead to exact
singularity in most models, it certainly reduces the
importance of the difference between the NET and
LET views.
Supply shocks provide an interesting litmus test
for deciding whether one is in the NET or LET camp.
Consider a sharp increase in commodity prices. In
many standard models, this shock tends to push
inflation up and push output below the efficient
level.13 Even if one leaves commodity prices out of
the inflation measure, there will be indirect upward
pressure on inflation. BLMP conclude that “a supply
shock that is great enough or that arises from some
unanticipated source may justify missing or changing a previously announced inflation target” (p. 35).
In our view, this conclusion puts BLMP squarely
in the LET camp. Mishkin (2002) makes similar
arguments.
One might hope that empirical evidence would
resolve this debate. The problem is that the distinction regards trade-offs along the efficient policy
frontier. Informally, the LET view suggests that at
the optimum the only way to reduce inflation variance is to raise gap variance. The NET view is that,
in the face of our profound ignorance, our best guess
is that any deviation from the policy of smoothing
inflation will increase both inflation and output
variance. That is, policy injects variance into the
economy with no expected benefit. Both sides agree
that many of the significant policy changes we find
in the data are movements toward the efficient
frontier. Such moves may result in improvements
in all aspects of performance.
In the remainder of the paper, we consider
economies that have all the features of macro common ground and that are consistent with the LET
view. Thus, much of our analysis will be of limited
relevance for NETers.
13

Included are models with explicit microeconomic foundations, such
as the one used by Aoki (2001).

122

J U LY / A U G U S T 2 0 0 4

REVIEW

4. DOES THE COMMUNICATION POLICY
OF THE ITF MAXIMIZE PUBLIC
UNDERSTANDING?
In this section, we begin our assessment of the
ITF with the simplest case. We set aside politicaleconomy problems and maintain the view that
nothing in the ITF constrains the central bank from
implementing the social optimum. Thus, there is no
question of whether the ITF delivers good policy.
That issue aside, the only remaining question is
whether the communication policy of the ITF constitutes a good implementation of transparency.
Our core requirements of the ITF give no details
about how transparency is to be achieved, so this
section also begins our filling in of the details of
the ITF. We begin by presenting some arguments in
favor of transparency, or as we put it, maximizing
public understanding.

4.1 Why Maximize Public Understanding?
In the recent past, few central banks would have
placed heavy emphasis on maximizing public understanding. Moreover, there is no general presumption
that increasing common knowledge in society
improves welfare. The transparency literature is rife
with examples where this is not the case, for example, Faust and Svensson (2001). Indeed, much of
the transparency literature can be viewed as a study
of when it is and is not optimal for the central bank
to surprise the public deliberately. This conventional transparency literature does not address
three arguments in favor of clear communication
that are stressed by ITF advocates and many other
commentators.
First, as Greenspan (2002, p. 6) states, “Openness
is an obligation of a central bank in a free and democratic society.” A great many conservative and liberal
economists have supported this view. Deliberately
surprising the public, even for its own good, is not
the proper role of a central bank, in this view.
The second reason for clarity is that, as Lucas
makes clear, what constitutes optimal policy is
inextricably linked with public expectations about
policy. The effects of a given policy action are not
even defined without a treatment of policy expectations. More recently, discussions of the liquidity
trap have reminded us of this point. The liquidity
trap case drives the point home because, under
certain assumptions, expanding the monetary base

FEDERAL R ESERVE BANK OF ST. LOUIS

in a liquidity trap has no direct effects on the
economy.14 Any effects result from changes in
expectations.
Lucas argues that, even away from the liquidity
trap, both the agent’s problem and the central bank’s
problem in practice are intractable unless the public
understands what the central bank is doing. An
assumption about public understanding of the future
course of policy is a precondition for coherent analysis of current policy.15
Accepting the role of expectations in the economy does not imply that central bank communication is important. Instead, Friedman and Lucas both
argued for very simple policies that would largely
obviate the need for communication. It is when we
accept the view that best-practice cannot at this time
be codified in a simply communicable way that continuing central bank explanation of actions becomes
essential.
The third argument in favor of clear communication is that it may alter incentives in a beneficial
way. Many variations of this idea have been studied
in the political-economy literature. We argue that ITF
advocates have a new channel in mind. This section
focuses mainly on more direct benefits of clear
communication; incentive effects are dealt with in
section 6.
It is useful to note that none of these three
reasons for clear communication has received much
emphasis in the transparency literature. These may
be the most important reasons for transparency in
practice, however.

4.2 Our Approach and Model
Analyzing communication policy is complicated
by the interaction between how policy is conducted
and how it is communicated. If communication
policy actually matters, then there is an interaction
between what one should do and what one should
say. Here we cut this knot by constructing an example in which we can unambiguously determine
what one should do. In particular, we examine a
simple model solved under the standard rationalexpectations assumption that all agents fully understand the model and policy. We give the central bank
the commitment technology to solve any politicaleconomy problems. In this case, the socially optimal
policy is unambiguous.
14

For one discussion of the liquidity trap situation and references to
many more, see Clouse et al. (2003).

15

A recent confirmation that central bank talk matters is provided by
Kohn and Sack (2003).

Faust and Henderson

We then assess whether the ITF communication
policy provides the most effective way to describe
the conduct of policy. If we added an uninformed
agent to the economy, would the ITF communication
policy be the best approach to bringing that agent
up to speed?
The model we employ has many standard features, and models like it have been used by supporters of the ITF, for example, Svensson (1997a). While
it is exceedingly simple, it embodies the common
ground described above. In our view, adding the
complexity of reality would only tend to magnify
the importance of the points we emphasize.
The model starts with the policymaker’s loss
function, which is the standard expected discounted
sum of period losses conditional on available
information:

(1)

 `

L t = 1 ε t  ∑ β j lt + j  ,
2  j =0


(

)

2

(

)

2

lt + j = π t + j − π ∗ + λ yt + j − y P − κ ,

where εt is the operator giving expectations conditional on time t information, ,t+j is the period loss
at time t+j, and 0<β<1 is the policymaker’s discount factor. The symbols yt+j and πt+j represent the
logarithms of output and gross inflation, respectively;
π* is the bliss value for inflation; y P is flexible-price
output, which we refer to as potential output; and
y*=y P+κ is the bliss level of output. We set κ ≠ 0
as in the time-consistency literature to allow for the
fact that the central bank may aim for output above
potential due to political pressure or for some reason
associated with economic distortions. The output
gap is defined as actual output minus potential
( yt+j – y P ).
The policymaker minimizes the loss function
subject to the Phillips curve,16
(2)

(
+α( y

)
)+ε

(

π t + j − π = φ π t + j −1 − π + (1 − φ )β π t + j +1|t − π
t+ j

−y

P

)

t+ j ,

where π– is the unconditional mean of inflation.
Deviations of inflation from its unconditional mean
at time t+j depend positively on both past and
16

Using Phillips curves that include both lagged and expected future
inflation is common practice; the exact specification in equation (2)
is used in Clarida, Galí, and Gertler (1999). One way of arriving at this
Phillips curve (2) is to assume that inflation rates are set in Calvo-type
contracts and that the inflation rates of agents who do not get to reset
prices in the current period are indexed to the unconditional mean
of inflation.

J U LY / A U G U S T 2 0 0 4

123

REVIEW

Faust and Henderson

expected future deviations, the output gap, and an
i.i.d. normal cost shock, εt. The symbol xt+j|t represents the expected value of x at time t+j conditional
on information available at time t.
The Phillips curve, (2), reflects two features of the
common ground. First, there is no trade-off between
mean inflation and any other mean or variance.17
This feature implies that the mean inflation rate can
be set independent of other considerations in the
model. In a more realistic model, there might be a
link between, say, the mean and variance of inflation,
but so long as the relation is generally positive this
does not change the argument that low inflation can
be chosen without regard to other goals.18
Second, there is a short-run trade-off between
inflation and the output gap, and as we shall see,
this trade-off is exploitable. The trade-off could be
made fuzzier in various ways, but doing so would
not alter the implications we emphasize.

4.3 Optimal Policy in a BackwardLooking Version
We use two special cases of the model to illustrate different features of interest.19 Here, we consider a backward-looking version of the model in
which there is no wedge between potential and
desired output (κ=0) and current inflation depends
on lagged inflation but not on expected future inflation (φ=1). Under these assumptions, the model
generates no inflation bias and, under optimal commitment policy, both the output gap and inflation
follow autoregressive processes20:

(

)

(3)

π t + j − π ∗ = Λ π t + j −1 − π ∗ + Λε t + j , 0 < Λ < 1

(4)

Λ −1
Λ −1
π t + j −1 − π ∗ −
εt + j ,
− yP =
α
α

yt + j

(

)

where the parameter Λ is defined in the appendix.
These processes have the following implications
that we will use in our discussion:
17

This property can be confirmed by taking the unconditional expectation of the Phillips curve.

18

Of course, there could be conflict among various low rates, as the
bliss points for the mean and variance need not coincide.

19

Analyzing the general model might be nicer in some respects, but
would be unduly complicated given our very limited ambitions. Both
Clarida, Galí, and Gertler (1999) and Svensson (2003) consider backwardand forward-looking versions separately.

20

All derivations are in the appendix.

124

J U LY / A U G U S T 2 0 0 4

1. Both inflation and the output gap are
covariance-stationary, Gaussian time-series
processes.
2. The unconditional expectation of inflation is
the target value, ε πt+j=π*.
3. Conditional inflation expectations are
described by πt+j|t=π*+Λ j(πt – π*).
4. There is an optimal balancing of output gap
and inflation variance.
Implications like these are very general given the
features of the common ground and the LET view.

4.4 Strengths of the ITF Communication
Framework: Transparency and Anchoring
Under the assumption that the central bank
implements the socially optimal policy as we have
derived it, we can now ask whether several usual
features of the ITF represent an effective way to
communicate best-practice. As stated in the introduction, the first goal of best-practice is to get mean
inflation right. In the model, we have that
(5)

lim π t + j|t = π ∗,
j →`

so it is clearly appropriate to announce a long-run
inflation target. Explanation of the behavior of inflation relative to the target is the centerpiece of ITF
communication policy. A primary objective of the
ITF is to anchor long-run inflation expectations, and
the policy leaves little room for misunderstanding
this objective. Thus, the ITF communication policy
is arguably extremely successful in communicating
about the first goal of monetary policy.

4.5 Room for Improvement:
The Balance of Multiple Goals
In this section, we argue that the primary shortcoming of the ITF communication policy is that it
does not explain clearly the roles and balance of
multiple goals. Indeed, we argue that the ITF as
implemented often involves elements that are literally inconsistent with best-practice policy and, in
any case, obfuscates some basic issues.
To begin discussion, we list some usual features
of the framework that we find problematic. It is
often a feature of the ITF, as advocated and practiced,
that one or more fixed horizons are associated with
the inflation target. This practice is literally inconsistent with optimization. For example, the results
above imply that

FEDERAL R ESERVE BANK OF ST. LOUIS

(6)

π t + j|t ≠ π ∗

j≥0

with probability 1, and that with certainty we will
face times, t, at which,
(7)

| π t + j|t − π ∗ | > ε > 0

j≥0

for any ε. That is, under best-practice, there will be
times when the expectation of inflation at any
horizon remains far from any target or target range.
Choosing a fixed horizon at which the inflation
forecast must be consistent with the target in some
sense can be thought of as an approximation to
optimization. In particular, under full optimization
we can pick a horizon h, a small probability ε, and
a margin of error θ such that
(8) pr(| π t + h|t − π ∗ |< θ ) = pr(π t + h|t ∈π ∗ ± θ ) = 1 − ε .
That is, at the horizon h, the forecast of inflation is
in a small neighborhood of π* most of the time. Thus,
choosing a fixed horizon for meeting the inflation
target seems like a sensible approximation. We take
up the costs and benefits of approximation in the
next section. For now, we note that choosing a fixed
horizon is not an accurate description of fully optimal policy.
Some central banks state target ranges for inflation. It is not clear how to interpret these ranges.
Does a central bank aim to be inside its announced
range all the time? Under best-practice, should it?
In our example, there is an interpretation of
the term “target range” that is consistent with bestpractice. As is clear from equation (8), the central
bank can view the target range as a confidence interval and relate the width of the range to the probability that inflation (or its forecast at the relevant
horizon) will be in the range.21 Under this interpretation, a target range is purely descriptive in that it
states that inflation will be within the range π*± θ
most of the time. This interpretation of the target
range is subtle, and we suspect not the predominant
one.
There is a contrasting interpretation that is not
consistent with best-practice. Under this interpretation, the central bank wants inflation to be inside
the range at all times, but control errors might cause
it to wander out at times. This interpretation is not
consistent with best-practice in the example. There
is no control error in our example—if there were,
21

The central bank can choose a probability, ε, and derive the width of
the range, θ, or pick a width and derive the probability.

Faust and Henderson

the θ associated with a given ε would be larger. Under
best-practice, the central bank deliberately sets
inflation outside any given interval at times. When
evaluating policy, no incident of inflation crossing
the boundary is evidence of central bank misbehavior; only excessive frequency of being outside the
interval constitutes such evidence.
A key test as to whether the range is properly
understood as a confidence interval is that under
best-practice excessive frequency of being inside
the range is also evidence of misbehavior. It should
seem equally natural to punish the central bank for
being inside the range too often as for being outside
the range. In the LET view, no matter how limited
one thinks the limited exploitability is, it remains
the case that excessive smoothness and excessive
volatility of inflation are equally costly at the margin
in equilibrium. As BLMP document, some of the
problems with target ranges we are pointing to have
been observed in practice.
Next we note that the long-run inflation goal is
often said to be preeminent in some sense in the ITF
framework.22 While the intention here may seem
clear enough, we do not understand what it means
formally.
Given the common ground we are accepting, it
is true that there is no trade-off between setting the
mean of inflation at π* and any other goal of policy.
There is no long-run trade-off; the mean of inflation
has no implications for other choices. Thus, the same
policy is obtained if inflation is the preeminent longrun goal or if setting the gap equal to zero is the
preeminent long-run goal.
Most crucially, we can arbitrarily rank the preeminence of long-run goals only if we are talking
strictly about the mean of inflation. Generally, the
preeminence statement is linked in some way to
price stability. To the extent that stability is interpreted in a natural way as having something to do
with the variability of prices and inflation, any statement of preeminence is the antithesis of the key
feature of optimal policy—the notion that optimization implies an optimal marginal rate of exchange
between stability of prices and stability of the gap.
22

For example, the Reserve Bank of New Zealand’s 2002 Policy Targets
Agreement states that “[i]n pursuing its price stability objective, the
Bank shall seek to avoid unnecessary instability in output, interest rates
and the exchange rate.” The Bank of England Act charges the bank
“(a) to maintain price stability, and (b) subject to that, to support the
economic policy of Her Majesty’s Government, including its objectives
for growth and employment.” There is similar language for the Swedish
National Bank, as confirmed by Heikensten and Vredin (2002).

J U LY / A U G U S T 2 0 0 4

125

REVIEW

Faust and Henderson

Here is the essence of our argument so far.
Monetary policy in the LET view involves conflicting
goals. The mean inflation goal may reside outside
this conflict, but any discussion of stability of prices
or inflation must inevitably raise issues of other goals.
We argue that several aspects of the ITF as practiced
do not provide a natural and straightforward framework for communicating this fact. We now consider
various ways other goals are accommodated.
One approach to balancing multiple goals is to
state a target range for inflation that is assumed to
give the central bank wiggle room to consider other
goals. In practice, things have arguably worked this
way. We return to the primary question of this section: Is wiggle room the most effective way to communicate optimization with multiple conflicting
goals? In our view, this way of communicating can
clearly work, but is not the height of pedagogy.23
Escape clauses are another alternative. Every
framework will surely need the equivalent of escape
clauses. There will be events sufficiently peculiar
from the standpoint of what was foreseeable at the
time the framework was conceived that briefly
abandoning the framework will be necessary. Still,
taking account of the role of other goals through
escape clauses is surely not fully transparent.
Finally, Svensson (1997a) has argued that in a
quadratic optimization framework like the one in
our simple example, we can view optimal policy as
targeting the forecast of inflation, with consideration
of the gap incorporated by allowing it to affect the
horizon at which one wants the forecast to hit the
target. Mishkin (2002) argues for this approach
and some ITF banks use this sort of rhetoric. This
approach is consistent with optimization. Formally,
it will be true under optimal policy that at each point
in time, t, there is a shortest horizon, h, such that
(9)

| π t + h|t − π ∗ | < θ

for a given θ. Thus, in each period an h could be
announced.
If we were teaching this optimization process
to an undergraduate or to the marginal agent added
to the model, is this the most natural way? We have
a linear-quadratic optimization with two conflicting
goals and one instrument. The two goals of optimization are to get the mean right and to balance vari23

Faust and Svensson (2001) present an example in which inflation
fluctuates narrowly around the optimum value, but due to lack of
transparency about the nature of other goals the economy is significantly more volatile than under full transparency.

126

J U LY / A U G U S T 2 0 0 4

ability of inflation against variability of the gap. It
seems strained, at best, to describe the optimization
process in terms of a target for one variable and
adjusting the horizon to take account of the other.
This description fundamentally obfuscates the
trade-off in question.
In this section, we have tried to make a simple
point that many people find obvious. The ITF
communication policy is tilted heavily toward
emphasis on stabilizing inflation. Several usual
features give inflation a role that is literally inconsistent with optimization in the LET perspective.
Thus, in our view the communication policy of the
ITF is not the best-practice way of maximizing
public understanding.

5. SIMPLIFICATION AND
APPROXIMATION
We have followed major advocates in interpreting
the ITF as allowing the central bank to follow the
socially optimal policy. Using this interpretation in
a conventional LET-view model, we find that a dissonance arises between policy and the standard
communication approach followed by the ITF. Perhaps we are being too literal: It may be that policy,
or the communication of policy, is deliberately
intended to be some sort of approximation of optimal behavior. These simplifications may be optimal in some broader perspective: Perhaps there is
some unmodeled simplicity constraint on either
communication or policy itself that we have not
captured. At some level, there surely are such constraints, so this possibility deserves serious treatment.

5.1 Simplicity-Constrained Policy
Perhaps policy behavior is subject to a simplicity
constraint that causes policymakers to follow ruleof-thumb-like policy. From where would such a
constraint arise? The standard justification is that it
arises from some need for ease of monitoring. Thus,
a bank with a severe credibility problem might find
that the credibility benefits of a rule that is trivial
to monitor outweigh the costs. Fixed exchange rates
are often justified in this way both in theory and
practice (Atkeson and Kehoe, 2002).24 The arguments
24

We are dealing here with the case in which the simplicity constraint
binds in the sense that the central bank deviates from the policy that
is best on standard macro-stabilization grounds. Thus, we are distinguishing this case from the one in which simple rules are best, even
from a pure stabilization standpoint. For example, Friedman argued
that a k percent rule is optimal due to our profound ignorance. Others
have argued that simple rules may be optimal, or nearly so, from the
standpoint of robustness (Levin and Williams, 2003).

FEDERAL R ESERVE BANK OF ST. LOUIS

for a simplicity constraint in extreme cases are
familiar.
Regarding the advanced economies that are
the focus here, we make two points. First, if the ITF
requires deviating from the optimal policy on economic grounds, then proper evaluation of the ITF
requires a clear statement of the deviations required.
In this case, we need to go back to the macromodel
horse races to evaluate the costs of the deviations
and attempt to weigh these costs against the benefits
of simplicity. Second, as we argue next, even when
banks assert that policy is constrained in this way, it
often turns out that only communication is actually
constrained.

5.2 Optimal Policy/SimplicityConstrained Communication
We have generally treated the problematic elements of the ITF as constraining communication,
not policy behavior. In practice, public communication requires some simplification, and as economists, we naturally think of simplification in terms
of an approximation that is adequate so long as
variables stay near some mean or steady-state values.
The problems with ITF communication listed above
will probably be minor so long as inflation and the
gap stay near the steady-state values.
We believe that use of a communication policy
simplified in this way is dangerous. As we get further
from the steady state, the appropriateness of the
simplified framework diminishes. Of course, since
these conditions are observed infrequently, uncertainty on the part of the public about the central
bank’s policy is greatest at these times. Further, the
conflict in society over the proper short-run policy
becomes more intense as we move away from the
steady state.25 Thus, the simplified communication
works best when it is least needed and tends to break
down when it is most needed.
If any ITF banks are following this course, they
are on a well-trodden path. Central banks have
regularly adopted rule-of-thumb communication
devices that function well during normal times and
then scrambled to wean the public from these rules
(or adjust and qualify the rules) when times became
more challenging.
Intermediate money targeting illustrates this
claim. It provided a framework for the conduct and
25

For example, for the quadratic loss function used in the model,
∂ (πt – π*)2/∂πt and ∂ ( yt – yP )2/∂πt both rise as πt and yt move from the
steady state.

Faust and Henderson

communication of policy. Although it was not necessarily presented this way at the outset, intermediate
targeting was a simplifying approach that was viewed
as ex ante suboptimal on stabilization grounds.26
Under such a system, there inevitably comes a time
when the best judgments about how to run policy
conflict with the direction dictated by the intermediate target.
The central bank must then choose between
running policy it believes to be suboptimal or running
policy inconsistent with the framework it typically
uses in communication. In practice, banks generally
chose the latter option. Thus, the Fed regularly
redefined the target, redefined the target variable,
and simply ignored the deviation of the target variable from target. Similarly, some have argued that
the Bundesbank was an implicit inflation targeter
and ignored the intermediate money target when it
appeared inconsistent with inflation objectives.27
We are not at all critical of this solution: These banks
probably made the right choice in deviating from
the communication policy rather than from best
policy.
This case illustrates that adopting simplified
communication approaches need not actually simplify anything.28 Such communication works fine
in easy times. In challenging times, a dissonance
arises between the simple communication framework and the course of policy, generating a certain
degree of turmoil and confusion.

5.3 Must the Public Have a Simple
Yardstick?
One virtue of the problematic ITF features that
we discuss is that they give the public a simple yardstick by which to judge policy. Given lexicographic
preferences over inflation and other goals, an inflation target range, and a fixed horizon, inflation targeting becomes very easy to monitor. One simply
checks whether the inflation forecast is at the target
at the specified horizon.
26

Intermediate targeting is inherently suboptimal so long as the word
intermediate is not superfluous. Svensson (1999) has derived the
conditions under which an intermediate target is “ideal,” and this, by
definition, is when there is no (observable implication of the) distinction
between the intermediate and ultimate goal. When intermediate money
targeting was adopted, no one claimed that money was “ideal.”

27

See, for example, Svensson (1999) and Romer and Romer (2000).

28

The Fed’s recent experience with the bias in the directive arguably
provides another example.

J U LY / A U G U S T 2 0 0 4

127

Faust and Henderson

The virtue of this yardstick is ease of use; the
problem is that it is the wrong yardstick. From
Heikensten and Vredin (2002), it seems that the
Swedish National Bank (hereafter Riksbank) may
have come closest to explicitly advocating that the
public think of policy using such a simple rule of
thumb. Recently policy pursued by the bank has
deviated from this rule of thumb, perhaps illustrating
to some extent the sort of communication problem
we raise.29
Despite such examples, it is explicit or implicit in
many discussions that ease of monitoring demands
that the public be given a yardstick for measuring
policy that is relatively straightforward to use. We
believe that Fed policy over the past 15 years provides
a counterexample. Arguably, one of the most notable
aspects of Federal Reserve policy in the Greenspan
era has been the fact that the Fed has resisted the
temptation to characterize policy in terms of some
simplified, and thereby inherently suboptimal,
framework. The Fed has demonstrated that one can
run policy with at least reasonable success without
placing constraints on policy or communication
that are thought ex ante to be suboptimal on economic grounds.
The Fed’s approach in this period is at times
viewed with alarm and/or suspicion. Svensson (2003)
argues that failure to adopt the ITF is a smokescreen
that allows the FOMC freedom to secretly change
its goals. Others argue that a concrete goal is essential for accountability.
These arguments may be correct, but they have
been selectively applied. The second major goal of
policy in the LET view is stabilizing the gap. It has
become conventional wisdom that the gap is sufficiently difficult to measure and that communicating
a concrete goal for any particular measure of the
gap would be problematic. This view is taken as
adequate justification for not reporting a concrete
goal for a gap measure. Let us set aside for a moment
the factual question of whether difficulties in measuring the gap are different in kind or only in degree
from those in measuring inflation.
Even acknowledging measurement problems,
one must surely echo Svensson in asking whether
these problems might be used as a smokescreen,
allowing a central bank to shift its preferences
about output stabilization.30 Further, one must ask
29

For a discussion of the Riksbank’s policy during this period, see
Sveriges Riksbank (2003).

30

Faust and Svensson (2001) show that even modest variations in this
regard can be costly.

128

J U LY / A U G U S T 2 0 0 4

REVIEW
how the central bank could possibly attain credibility
and accountability on the gap goals without a concrete gap goal. These issues are no less pressing in
the case of real stability than in the case of inflation
stability.
In practice, we suspect that ITF advocates are
comfortable with the view that credibility and
accountability regarding real stability responsibilities
can be attained through vigorous central bank communication. By the same token, we argue that it is,
at the very least, an open question whether accountability and credibility regarding inflation necessitate
adopting a simple yardstick that is suboptimal in
the sense we have been describing.

6. POLITICAL-ECONOMY PROBLEMS
AND THE ITF COMMUNICATION
POLICY
ITF advocates contend that central bank communication can solve political-economy problems.
In our view, this is the least well-analyzed claim of
ITF advocates. In this section we describe the communication channel emphasized by ITFers and show
how existing tools can be employed to analyze it.
We ultimately conclude that use of this channel can
play a role in getting the mean of inflation right, but
seems as likely to complicate as to facilitate achieving
the appropriate balance between inflation and output stability.

6.1 The Communication Channel
The basic idea behind the ITF communication
channel is that people dislike exposure of their
intentional trickery, honest mistakes, or incompetence. If this is so, then public promises carry their
own enforcement mechanism based on policymaker
aversion to criticism. The ITF is designed to make
better use of this channel by requiring public statement of goals and then public reports about progress
on the goals. As BLMP (p. 25) argue,
To the extent that the central bank governors
dislike admitting publicly that they may miss
their long-run inflation targets (or, alternatively, to the extent that they dislike having
their inflation projections criticized as biased
or manipulated), the existence of an inflationtargeting framework provides an incentive
for the central bank to limit its short-run
opportunism.

FEDERAL R ESERVE BANK OF ST. LOUIS

Svensson (1999, p. 663) is more emphatic:
I believe it fair to say that never before in
monetary history has an incentive system
been set up with such strong incentives for
optimal monetary policy decisions.
In more formal terms, the communication
channel invokes terms in the central banker loss
function—associated with, say, honesty and aversion
to criticism—that have often been ignored.31
The ITFers’ argument here seems consonant
with two alternative views about political-economy
problems. Blinder (1998) and McCallum (1997) argue
that the time-consistency literature simply misses
the point. According to McCallum, knowing about
the commitment policy, central bankers would “just
do it.” The ITF communication policy might be seen
as an attempt to increase the probability of this outcome. Friedman has a different take on central
bankers’ loss functions:
From revealed preference [as revealed in
central bank communication], I suspect that
by far and away the two most important
variables in their [Federal Reserve policymakers’] loss function are avoiding accountability on the one hand and achieving public
prestige on the other. (quoted in Fischer,
1990, footnote 52)
Whether or not one subscribes to such an
uncomplimentary view, why not design a framework
to constructively exploit motives such as a desire
for prestige?
We readily accept the ITF premise that the threat
of public criticism affects the incentives of the central
bank and thereby the course of policy. We take the
view (perhaps following Blinder and McCallum) that,
in normal times and with first-rate policymakers,
this channel may not be of great importance. These
31

The conventional literature has examined several channels through
which talk by the central bank could alter equilibrium outcomes. For
example, it could lead to the cheap-talk equilibrium of Stein (1989),
facilitate coordination on the best of the many equilibria of a monetary
policy game as shown by Barro and Gordon (1983), or beneficially
expand the set of equilibria as illustrated by Atkeson and Kehoe (2002).
ITF advocates have in mind something much simpler—although formalization could involve elements like those discussed above. The
communication channel as described here has a family resemblance
to what Barro and Gordon called reputational equilibria in which the
public might raise its expectation of future inflation in order to “punish”
the misdeeds of the central bank. Technically, the important distinctions
here are that the disutility from the punishment falls directly on the
central bankers, involves no costs to the public, and is automatically
attached to failure to deliver on “promises.”

Faust and Henderson

policymakers will “just do” the right thing, as they
see it, largely independent of public accolade or
criticism. In the spirit of preparing the roof for rainy
days, however, we consider the case when a weaker
or more political board is in place.

6.2 Solving Political-Economy Problems
Using Special Loss Functions
The communication channel involves terms in
the loss function representing aversion to criticism
that are usually neglected. Fortunately, the literature
provides tools for studying solutions to politicaleconomy problems using special loss functions.
Rogoff (1985) considers simply picking a “conservative” central banker, one with a loss function that
embodies greater aversion to inflation than the true
social loss function. This approach generates a
trade-off: Excess aversion to inflation lowers mean
inflation, but it causes inflation to be smoother and
the output gap to be more variable than is optimal.
Melitz (1988) and Obstfeld (1996) make use of terms
in the loss function embodying “political costs”
associated with breaking a pledge to keep the
exchange rate fixed. They, too, generate a trade-off
between reducing inflation bias and achieving stabilization objectives. Calling attention to this possible
trade-off is one of the main contributions of the timeconsistency literature.32 Canzoneri, Nolan, and Yates
(1997) provide examples in which taking advantage
of “political cost” terms generates more complicated
trade-offs. These papers all consider intuitively
appealing, but ad hoc, loss functions.
In contrast, Walsh (1995) and Persson and
Tabellini (1993) show how one can derive a loss function that completely eliminates the inflation bias
problem without introducing stabilization costs. They
discuss how this structure of loss might be induced
using performance contracts for policymakers.
More recently, Lockwood, Miller, and Zhang (1995),
Svensson (1997b), and Svensson and Woodford
(forthcoming) construct loss functions that eliminate
both inflation and stabilization biases in models
like ours. The basic approach in all these studies is
to amend the policymaker loss function in such a
way that the implied first-order conditions under
discretion give rise to the same policy as under
commitment.
32

Canzoneri (1985) also calls attention to such a trade-off. In his paper,
the trade-off arises because of the imposition of an additional constraint
on the policymaker, a requirement to achieve an average value for the
money supply, not because of a special policymaker loss function.

J U LY / A U G U S T 2 0 0 4

129

REVIEW

Faust and Henderson

6.3 A Formal Example
Here we use a forward-looking version of our
simple model to illustrate how amending the loss
function can eliminate inflation bias and stabilization bias. In this version, we assume that current
inflation depends on expected future, but not lagged,
inflation (φ=0) and that target output exceeds potential (κ >0). For simplicity, we assume that there is a
single random shock in period t (εt ≠ 0, εj=0, j ≠ t).
Under commitment, the policymaker can affect
inflation at t and expected inflation in all future
periods. Therefore, it can smooth adjustment to the
shock over multiple periods. In future periods, the
policymaker has an incentive to renege but is locked
in by commitment. However, under discretion, when
the shock hits at t, the policymaker cannot have the
desired effect on inflation expectations in future
periods because it cannot be relied upon to ratify
those expectations. Therefore, inflation returns to
its unconditional mean in period t+1 and remains
there, so adjustment to the shock is smoothed less
effectively. The formal solutions are presented in
Table 1.33 The commitment policy we consider is
commonly referred to as the full-commitment policy
(or the solution to the Ramsey problem): It is the
unconstrained optimum given that the policymaker
can commit to the chosen policy.
Our model exhibits the classic inflation bias
under discretion. Ignoring any shocks, at the optimum inflation rate of π*, the bank has an incentive
to surprise the public by increasing inflation in an
attempt to stimulate output. Inflation is increased
to the point at which the marginal cost of additional
inflation just offsets the marginal benefit of attempting to raise output above potential toward the target.
To see the inflation bias in this case, set εt equal to
zero in the discretion solutions in equations (T1.1)
and (T1.2) and assume that κ >0. Under discretion,
inflation is constant and exceeds π* by the standard
inflation bias, λκ /α.
In contrast, under commitment, inflation is time
varying. It approaches π* from above and is always
less than the positive inflation under discretion.
Output is also time varying. It approaches y P from
above and is always below y*. It is optimal to have
inflation above π* in order to raise output above y P
for all finite j, but optimal inflation and output
must decline over time in order to be consistent
with the Phillips curve.
33

As before, all derivations are in the appendix.

130

J U LY / A U G U S T 2 0 0 4

The model also exhibits stabilization bias under
discretion. A useful measure of stabilization bias is
the part of the extra loss from discretion relative to
commitment that results from the existence of the
shock. To consider stabilization bias, set κ equal to
zero in all the solutions in Table 1 and assume that
there is a positive cost shock in period t (εt>0) and
no shock in any other period. As might be expected,
under discretion the optimal response to the cost
shock in period t involves some increase in inflation
and some reduction in output in period t and no
response in any later period because there are no
shocks then.
Under commitment, the optimal responses of
inflation and output in period t are damped relative
to those under discretion. There are reductions in
both inflation and output in period t+1 and in every
period thereafter, with the responses approaching
zero from below as j approaches infinity, in order
to be consistent with the Phillips curve. The Phillips
curve in period t implies that reducing inflation in
period t+1 partially offsets the upward pressure
on inflation resulting from the shock. Therefore, it
is possible to damp the responses of both inflation
and output in period t. The benefit in period t more
than offsets the losses in all future periods because,
with a quadratic loss function, the first small movements away from the optimum in future periods
cause negligible increases in loss.
We can attain the full-commitment solution
under discretion if the policymaker is given the
amended loss function
(10)
Lt =

εt

(

)

(

 π − π∗ 2 + λ y − yP −κ
t+ j
 t+ j
1
∑ β j
2 j = 0 − α 2 + λ − λκ  π
t+ j
α 
 
`

) 
2





λκ


−  Cπκ ,t
+ Cπε ,t ε t  π t


α
`
λκ


− ∑ β j L j  Cπκ ,t +1
− Cπε , t +1 ε t π t +1+ j .


α
j=0
Loosely speaking, one subtracts out terms in the
discretion first-order condition that lead to the
inflation bias and adds in terms relating to the timevarying and state-contingent optimal policy. In particular, stabilization bias can be eliminated only by
subtracting terms that are time varying and depend
on the size of the cost shock.

FEDERAL R ESERVE BANK OF ST. LOUIS

Faust and Henderson

TABLE 1
Discretion Solutions
λκ
+ Dπε ,t ε t ,
α

(T.1)

π tD = π ∗ +

(T.2)

π tD+ j|t = π ∗ +

λκ
,
α

ytD = y P − D εy,t ε t ,

ytD+ j|t = y P,

j = 1, 2,...

Commitment Solutions
λκ
+ Cπε ,t ε t ,
α

(T.3)

π tC = π ∗ + Cπκ ,t

0 < Cπκ ,t < 1 = Dπκ ,

(T.4)

ytC = y P + Cκy,t κ − C εy,t ε t

(T.5)

λκ


π tC+ j +1|t = π ∗ + Λ j  Cπκ ,t +1
− Cπε ,t +1ε t  ,


α

(T.6)

0 < Cπκ , t +1 < Cπκ ,t < Dπκ ,

(T.7)

λκ


ytC+ j +1|t = y P + Λ j  Cκy,t +1
− C εy,t +1ε t  ,


α

(T.8)

Dκy, t +1 = 0 < Cκy, t +1,

0 < Cπε ,t < Dπε ,t

Dκy,t = 0 < Cκy,t < 1,

0 < C εy,t < D εy,t

j = 0,1,...

Dπε , t +1 = 0 < Cπε , t +1 < Cπε ,t
j = 0,1,...

D εy, t +1 = 0 < C εy, t +1

6.4 Likely Effectiveness of the
Communication Channel
Our example illustrates how political-economy
problems can be avoided if only the policymaker’s
loss function differs from the social loss function
in the proper way. It also shows that, even in a
stripped down model, the required special terms
are somewhat complicated and vary with the state
of the economy. In this very simple model, we might,
however, imagine writing an incentive contract to
achieve optimality. Of course, it is a feature of the
common ground that the actual economy is so complicated that it is impossible to codify such a contract.
ITFers argue that we can achieve much of the
desired effect by exploiting policymaker aversion
to criticism. However, to achieve this effect on the
first-order conditions, these aversion-to-criticism
terms would have to take a very particular, statedependent form. We have seen no argument as to
why aversion to criticism would work in the intended

manner. Indeed, we have difficulty imagining how
to form such an argument.
More generally, we know of no literature supporting the view that some combination of public
promises, maximal scrutiny, and the threat of public
criticism is an uncontroversial recipe for optimal
public policy. Suppose we accept that weak policymakers will be swayed by criticism. Given the skewing of communication in the ITF, it strikes us that a
weak policymaker may find it safest to excessively
smooth inflation. That is, it might literally follow its
pronouncements. Both this view and the contrary
view, however, are highly speculative given the
nature of the mechanism.
Despite our reticence to draw strong conclusions
either way, we offer two comments. First, we find it
plausible that stating a long-run inflation goal and
communicating regularly about it will raise the
chance that policy hits the long-run goal for the
reasons cited by the ITF. Second, the ITF seems as
likely to complicate as to facilitate achieving a proper
balance of multiple goals by a weak policymaker.
J U LY / A U G U S T 2 0 0 4

131

REVIEW

Faust and Henderson

7. COMPLICATING FACTORS LARGELY
MISSING FROM MODELS

sion about policy. Goodfriend (1986) lays out various
forms of this argument including the following:

Up to now, we have been analyzing the ITF from
a relatively abstract perspective. As preparation for
making some constructive suggestions based on our
analysis, we discuss some real-world complications
that are largely missing from our analysis and most
other formal work.

In this view, secrecy could confer a social
benefit because it makes consensus politics
work more smoothly and with less cost.

7.1 Strategic Skewing and Transparency
Beyond clarity, strategic skewing may be an
essential component of effective public communication. One of the most famous principles of strategic
skewing in the folk wisdom of central banking is that
central banks should “do what they do, but only
talk about inflation.” Alan Blinder contravened this
principle, the story goes, in the famous Jackson Hole
imbroglio, perhaps confirming its wisdom.
While most ITF advocates would vigorously
dispute it, the ITF might be viewed as an application
of this folk wisdom. Without the folk wisdom, it is
difficult to imagine why a policy of optimization with
multiple conflicting goals would be called “inflation
targeting.” Calling reports on all aspects of policy
“inflation reports” is an analogous misnomer. The
folk wisdom would also justify discussing goals
other than inflation only as they affect the horizon
over which one intends to hit the inflation target.
The folk wisdom might actually be wisdom. If
these topics are too sensitive to discuss, then we
should stop praising the ITF and other central banks
for their commitment to transparency; instead we
should lament the fact that central banks cannot
publicly discuss the pursuit of their multiple mandates. Further, the practical art of strategic skewing is well-studied, for example, by public relations
experts. Economists have no special claim to
expertise in optimal skewing.

The role of concerns like this in determining the
current structure of the Federal Reserve is reviewed
in Faust (1996).
We think it is almost certainly a grave error to
trivialize such concerns. We cannot contribute much
to their analysis, however. In our view proper treatment requires bringing in expertise from areas
beyond economics. We stick to reviewing the more
directly economic merits of transparency.

7.3 Multiple Decisionmakers
In many countries a board is charged with
making policy. In much of the theory we have been
reviewing, the fact of multiple decisionmakers is
inessential. The framers of the Federal Reserve, however, saw the composition of the board as an essential aspect of their response to political-economy
problems. For example, Warburg (1930, p. 773)
argues that a
formula had to be found by means of which
these two elements [big business and politicians] would be called upon to balance one
another.
As is often the case with responses to difficult
political design problems, the framers came up with
a mish-mash solution.34 As a bottom line, they
decided on a particular weighting of interests on
the FOMC. Congressman Henry Steagall (1935,
p. 13706) summarized the result:
[U]nder the bill embodied in the conference
report the board [that is, the FOMC] will
stand 5 to 7 giving the people of the country,
as contradistinguished from private banking
interest, control by a vote of 7 to 5 instead
of by a vote of 3 to 2 [as proposed in the
Senate].

7.2 Maximal Transparency and
Deliberation
Unlimited transparency may be inconsistent
with optimal deliberation. As Greenspan (2002, p. 5)
puts it
The undeniable, though regrettable, fact is
that the most effective policymaking is done
outside the immediate glare of the press.
More generally, transparency could, depending on
the social environment, generate inefficient dissen132

J U LY / A U G U S T 2 0 0 4

34

For example, there are 12 votes on the FOMC; five presidents vote; the
president of the New York Fed and either the Chicago or Cleveland Fed
president always vote. The Reserve Bank presidents are nominated
by the boards of their respective banks and confirmed by the Federal
Reserve Board. The nominating boards are composed of nine directors,
six chosen by district bankers (three representing district bankers and
three representing general district interests), and three chosen by the
Federal Reserve Board. The seven governors are nominated by the
President of the United States with due regard to a fair representation
of the financial, agricultural, industrial, and commercial interests.

FEDERAL R ESERVE BANK OF ST. LOUIS

Multiple heterogeneous policymakers clearly
pose a practical problem for transparent communication about the goals of policy, the rationale for
policy, and the causes of past policy mistakes and
successes. There seems to be disagreement about
the magnitude of this problem, however. Some contend that there are not many differences of opinion
about either the appropriate loss function for policy
or about how the economy works. Others are not
so sanguine.
If there are substantial disagreements over the
loss function and/or the workings of the economy,
one response would be to require multi-stage decisionmaking. First, the board agrees on goals of policy.
Next, taking the goals as given, the board agrees on
the model of the economy. Finally, the board makes
policy taking the goals and model as given.
This approach has the convenient feature of
making the policy process, for purposes of analysis,
look rather like the simple single-decisionmaker
problem. In some discussion, failure to agree first
on goals of policy or the model sometimes seems
to be taken as prima facia evidence of inefficient
behavior by the policymaking board. This view
trivializes the analysis of public decisionmaking.
There is no theorem of public decisionmaking
stating that the multistage decisionmaking approach
is good for society. One can write examples in which
multistage decisionmaking is or is not efficient, but
theoretical examples probably do not get to the heart
of the matter. Imagine a monetary policy board
populated by astute, public-spirited, policy-oriented
economists—epitomized perhaps by James Tobin
and Milton Friedman. The multistage approach
would require that they agree first on goals, next on
the model, and only then consider policy options,
given those goals and model. In an alternative
approach, we could simply charge them with agreeing on and implementing policy. One suspects that
the multistage approach may not even be feasible
in practice. There is at least room to differ regarding
which approach would lead to better policy.

7.4 Measurement of Inflation and the
Gap
Choosing a measure of inflation and an appropriate long-run inflation goal presents practical
problems that may not have been fully appreciated
during periods of high inflation. In 1980, any low
number seemed like a good thing. Fortunately, gone
are days when proponents of low inflation could
simply assert, “Zero is a nice round number.”

Faust and Henderson

There is now general agreement that most, if
not all, price indices exhibit non-negligible quality
biases. Further, for various reasons to do with quality
bias and composition, different indices often behave
quite differently for substantial periods of time. Thus,
the issue of which index to focus on may be of
some importance.35
Many economists inside and outside the ITF
camp have also concluded that the inflation goal
should be set high enough to limit the probability
of hitting the zero lower bound on nominal interest
rates. Thus, they argue for allowing a non-negligible
stabilization buffer that exceeds proposed allowances for quality bias.36 Exactly how large a buffer
to allow is a technical question involving the costs
of moderate inflation, the costs of being mired in a
recession, and the responsiveness of the economy
to changes in the policy rate. The answer may change
over time as the economy changes.37 In our view,
the fact that choosing the appropriate target is a
technical question and the possibility that a good
initial answer may not be found argue strongly for
keeping the choice of the inflation target in the hands
of central banks and for revisiting it periodically.
Obviously, ITF banks believe that the benefits
of a concrete target outweigh any costs that might
be generated by these measurement issues. Some
current and past U.S. policymakers agree—for example, Bernanke (BLMP) and Meyer (2002). Gramlich
(2003) speculates that announcing a long-run range
for inflation might increase transparency without
unduly limiting flexibility. In contrast, Greenspan
(2002, p. 6) argues that
For all these conceptual uncertainties and
measurement problems, a specific numerical
inflation target would represent an unhelpful
and false precision.
Greenspan argues, loosely speaking, that the
Fed is striving for price stability properly measured
and that achieving this goal does not correspond
reliably to hitting a long-run target for any particular
index. We do not attempt to resolve this empirical
dispute in this paper, but discuss some of its implications in the final section.
35

Recent discussion of the issues regarding changing the relevant
index in the United Kingdom (HM Treasury, 2003) and of the role of
energy prices in Sweden (Sveriges Riksbank, 2003) are a reminder
that the choice of price index can have important implications.

36

For a particularly eloquent statement of the case for a stabilization
buffer, see Phelps (1972, p. 210).

37

Henderson (2004) puts forward views similar to those expressed here.

J U LY / A U G U S T 2 0 0 4

133

REVIEW

Faust and Henderson

It is generally believed that the measurement
issues regarding the gap are much more serious.
Here, in discussing the real world, we are moving
beyond our model and using gap as a short-hand for
the relevant measure of economic slack. In practice,
there may be many relevant gaps, and none is easy
to measure. Indeed, one version of the NET view is
that these problems are so overwhelming that the
gap should not be part of policymaking.38
In the LET view, these measurement issues
may be immense and, hence, exploitability may be
minimal. By definition, however, the LETer believes
that we can monitor the economy sufficiently well
to attain some beneficial exploitation of the shortrun trade-off. If this is so, there is also some way to
communicate this information to the public. That
is, there should be a heavy presumption against the
view caricatured by Karl Brunner that central banking is an inaccessible art, and that the
esoteric nature of the art is moreover
revealed by an inherent impossibility to
articulate its insights in explicit and intelligible words and sentences. (as quoted in
Goodfriend, 1986)
Arguably, to the extent that there is less controversy regarding the measurement of inflation than
the gap, careful and thorough communication about
the gap is more essential. Under this view, the emphasis in the communication strategy of the ITF is
misplaced.

8. CONCLUSIONS
In section 2, we state that two core requirements
of the ITF are a long-run inflation target, and transparency. The first only constrains central bank policy
in the long run; the second constrains talk, not
actions. Based on these requirements alone, it would
be difficult to understand the passionate debate
over the ITF.
In filling out the description of the ITF, we have
come to believe that some of the passion comes from
the following characterization. Inflation targeting
has been portrayed as a snug-fitting garment that is
a great improvement over pure discretion. The latter
is caricatured as a seat of the pants approach. Less
colorfully, inflation targeting is depicted as constrained discretion.39

From one perspective, this characterization is
quite reassuring. Purely discretionary, seat-of-thepants policymaking sounds a bit risky. Given that
monetary policy is made in a complicated world and
is subject to conflicting pressures in society over
what policy is best, there is rightfully something comforting about the notion of a snug-fitting garment.
From another perspective the characterization
is distressing. Central banks have often put on snugfitting garments—for example, fixed exchange rates
and money targeting. Historically, these garments
have proven uncomfortable; they have regularly
split at the seams.
Which perspective is correct? ITF advocates
tell us there is nothing to fear from the snug-fitting
garment: The constraints in constrained discretion
do not constrain the central bank from pursuing
optimal behavior. While there is nothing, in principle,
wrong with this claim,40 we remain uncertain about
the basis of this claim. Going beyond the core
requirements, in practice ITF imposes many requirements—for example, fixed horizons, target ranges,
lexicographic preferences regarding price stability.
These requirements might be viewed as constraining
policy, but they are inconsistent with optimization
under conventional views about the economy. We
have treated them not as constraints on policy behavior, but merely as features of the ITF communication
policy that generate dissonance between how the
banks talk and how they act. We have not identified
requirements of the ITF that constrain use of policymaking discretion, while remaining consistent with
optimization.
We acknowledge that, despite their costs, central
banks might find simple yardsticks or rules of thumb
useful in guiding the communication of and/or conduct of policy. It is an open question whether such
simplifications yield net benefits in any particular
context. Until such constraints are better justified,
they are not, in our view, clearly part of best-practice
monetary policy.

8.1 Constructive Suggestions
Our review of the ITF leads to several constructive suggestions regarding best-practice, whether
inside or outside the ITF. In short form, these amount
to a guide for implementing the core requirements.
40

38

See, for example, Orphanides (2003a).

39

This characterization can be found in many places, including BLMP.

134

J U LY / A U G U S T 2 0 0 4

For example, “attain the social optimum” could be viewed as a constraint that imposes no costs. More generally, of course, constraints
on behavior off the desired equilibrium path have no cost of the sort
we are discussing and may help in selecting from among equilibria.

FEDERAL R ESERVE BANK OF ST. LOUIS

Suggestion 1. Central banks should state a clear
long-run inflation goal. No range or fixed horizon
should be given.41 If no numeric target is given, clear
countervailing interests should be stated, and effort
should be made to reduce uncertainty regarding the
long-run goal. The value of this goal in meeting all
long-run goals of policy should be stressed.
In order to be consistent with this suggestion, ITF
central banks would have to modify the characterization of their long-run inflation goals in many cases.
The Fed regularly communicates its strong commitment to the principle of price stability but has not
stated an explicit target. As justification for not stating an explicit target, various policymakers have
cited problems due to measurement issues, politicaleconomy considerations, and the existence of
multiple decisionmakers. A key question facing the
Fed is whether with careful statement and qualification, it might obtain some of the benefits of a more
concrete goal while minimizing the costs.
Suggestion 2. Central banks should communicate in a balanced way about the objectives driving
short-run policy. If these objectives are seen as conflicting due to the structure of the economy, this
viewpoint should be made clear. To the extent that
other goals are more difficult to quantify than the
inflation stability goal, the need for clear reporting
is heightened and banks should strive to find ways
to communicate about these goals effectively.
For ITF central banks, adopting suggestion 2
would involve changing some language that is literally inconsistent with optimization in the LET view.
It would also require providing more complete communication about the role of goals other than inflation in policy. It is a matter of perspective whether
these are viewed as small or large changes. They may
well be minor tweaks on the idealized ITF framework
defined by Svensson.
If transparency is truly a goal in central banking,
then the area with the greatest room left for improvement at most central banks is communicating the
roles of goals other than inflation. For ITF banks
and others, it would greatly clarify issues if three
questions were answered: Is some notion of real
stability an objective of the central bank? Does the
bank take the NET or LET view? If it is a LETer, how
will the trade-off be managed? At a general level, ITF
central banks and others have done quite well in
answering the first of these questions but less well
in answering the latter two.
41

Certainly no fixed horizon shorter than a business cycle.

Faust and Henderson

Answering the second question—are you a NETer
or LETer?—is relatively straightforward. Answering
the third question requires successfully characterizing how the trade-off will be managed. This is a
complex task that all LET central banks must face.
We have no special insights on this topic. We believe,
however, that clarity on goals and the NET/LET distinction is an important first step. It brings real stability into the conversation in its proper role. The focus
on inflation has led to valuable innovation in communication; putting real stability on the table can
promote similar innovation on the real side. The
Reserve Bank of New Zealand and the Bank of
Norway have begun this experimentation process.42
Suggestion 3. If best-practice policy is complicated, the totality of central bank communication
should reflect that complexity.
This suggestion is based on the distinction
between two dictionary senses of transparent. The
first, and the one generally applied throughout the
central banking literature, is that transparent means
“frank, open, candid.” However, sometimes the idea
surfaces that central bank communication should
be transparent in the sense of being, “easily seen
through, recognized, [or] understood.”43 If bestpractice dictates complicated policy, then communication that is frank, open, and candid will probably
not be easily understood.
In practice, central banks probably need to
have a multi-layered approach to communication
with differing levels of complexity. However, there
should be readily available information that makes
it possible for a reasonably tenacious and intelligent
person to understand policy in its complexity.
A review of the material on the web sites of
several ITF banks suggests several useful examples
of this multi-layered approach. Setting aside our
particular criticisms about overemphasis of inflation,
the monetary policy reports of ITF banks represent
an extremely valuable aid in understanding the
function of policy. These banks generally commission and publish outside reviews of the reports; for
this and other reasons, their content is steadily
improving.
Assessing the communication of the Fed is more
difficult. It is certainly true that there is no one
source of information that brings together the information represented in the best of the ITF monetary
42

For example, both include a forecast for the gap in their inflation
reports.

43

These definitions come from the Oxford English Dictionary, definitions
2a and 2b of transparent, respectively.

J U LY / A U G U S T 2 0 0 4

135

REVIEW

Faust and Henderson

policy reports. Due to the sheer bulk of reports to
Congress, testimony and speeches by FOMC members, and publications by the Board and 12 member
banks, it would be a daunting (and unenviable) task
to confirm reliably which topics are covered and
which are not. A key issue facing the Fed in vigorously pursuing effective transparency is whether
this material can be more concisely packaged and
delivered, while respecting the diverse committee
structure of the FOMC.
Suggestion 4. Central banks should strive to
communicate clearly the likely course of policy. If
forecasts are part of this process, the relationship
between the forecasts and the future course of policy
should be explained.
If policy cannot be codified in a simple rule, as
we have assumed, then helping the public understand the likely course of policy is an essential part
of effective policymaking. Central bank forecasts
are a central feature of the monetary policy reports
of ITF banks. While such forecasts can and do play
many roles,44 determining the proper role of forecasts in shaping policy expectations is more complex
than usually recognized.45 Under standard practice,
the forecasts are of unclear value in understanding
the course of policy.
For example, it is standard practice to report a
forecast for output and inflation conditioned on some
counterfactual path for policy such as a constant
policy rate. These forecasts are generally depicted
as judgmental, and the reader cannot know the
model in more than general terms. It is very difficult
to see how such a forecast has any marginal value
in predicting the course of policy, beyond the predictive power of the standard data available to the
public. For example, if the central bank optimizes
in the LET view, then the public certainly cannot
deduce that a conditional forecast of inflation above
target means policy will be tightened. Further, releasing a forecast for output growth (which is only tenuously related to the gap or other relevant notions
of slack) provides little help. Only if there is a known
mapping from the conditional forecast to policy is
that forecast of clear use.46
44

45

46

We note that there may be public good benefits from releasing a central
bank forecast that are independent of whether the forecast reveals
something about central bank intentions. We are setting those aside.
Any such benefits may be substantially reduced if the forecast is a
conditional one, however.
Anyone who doubts this claim is referred to the excellent work of
Svensson and Woodford (2003) on this relationship.
Leeper (2003) makes these and several other points about the use of
forecasts.

136

J U LY / A U G U S T 2 0 0 4

In contrast, an unconditional forecast of goal
variables and the policy rate would shed a good
deal of light on policy. Given that the public already
has its own unconditional forecast of the economy
and policy, the public can see if the two forecasts
differ and, by comparing the policy rate forecasts,
make an attempt to deduce whether those differences stem from different views of the future path
of policy or from different views of other aspects
of the economy.47
Of course, central banks have historically been
very wary of providing direct information about the
future course of policy. The political countervailing
interests may be overwhelming. Absent direct information about the future course of policy, a more
thorough discussion of the link between the forecast
and future policy would be useful.48

8.2 Summing Up
Overall, our four suggestions for best-practice
reflect the views that central banks should have clear,
though possibly conflicting, goals and should aspire
to maximize public understanding of policy. Advocates and practitioners of the ITF deserve great credit
for their many contributions to clear goal setting
and communication by central banks. Our suggested
approach differs from the standard ITF approach
in that we more strongly reject the folk wisdom of
central banking that all communication should be
couched in terms of inflation. Further, we do not
favor the use of concrete but inherently suboptimal
yardsticks for measuring central bank performance.
We advocate a more “candid and open” approach
to transparency.

REFERENCES
Alesina, Albert and Rosenthal, Howard. Partisan Politics,
Divided Government and the Economy. New York:
Cambridge University Press, 1995.
Aoki, Kosuke. “Optimal Monetary Policy Responses to
Relative Price Changes.” Journal of Monetary Economics,
August 2001, 48(1), pp. 55-80.
47

In addition to its inflation forecast, the Reserve Bank of New Zealand
releases forecasts of both the policy rate and the gap. To our knowledge
it is the only major ITF central bank that follows this practice. Svensson
(2001) has praised this practice and argued that others should follow it.

48

As discussed in subsection 5.3, the Riksbank has provided a rule of
thumb linking their forecast to future policy. This rule, if followed,
would make the forecast more useful, but is inconsistent with
optimization.

FEDERAL R ESERVE BANK OF ST. LOUIS

Faust and Henderson

Atkeson, Andrew and Kehoe, Patrick J. “Exchange-Rate
Based vs. Money-Based Monetary Policy: The Advantage
of Transparency.” Working paper, Federal Reserve Bank
of Minneapolis, 2002.

Faust, Jon, and Svensson, Lars E.O. “Transparency and
Credibility: Monetary Policy with Unobservable Goals.”
International Economic Review, May 2001, 42(2), pp.
369-97.

Barro, Robert J. and Gordon, David B. “A Positive Theory of
Monetary Policy in a Natural Rate Model.” Journal of
Political Economy, August 1983, 91(4), pp. 589-610.

Fischer, Stanley. “Rules Versus Discretion in Monetary
Policy,” in B.M. Friedman and F.H. Hahn, eds., Handbook
of Monetary Economics. Volume 2. New York: North
Holland, 1990, pp. 1155-84.

Bernanke, Ben S.; Laubach, Thomas; Mishkin, Frederic S.
and Posen, Adam S. Inflation Targeting: Lessons from the
International Experience. Princeton: Princeton University
Press, 1999.
Bernanke, Ben S. and Mishkin, Frederic S. “Inflation Targeting:
A New Framework for Monetary Policy?” Journal of
Economic Perspectives, Spring 1997, 11(2), pp. 97-116.
Blinder, Alan S. Central Banking in Theory and Practice.
Cambridge, MA: MIT Press, 1998.
Canzoneri, Matthew B. “Monetary Policy Games and the
Role of Private Information.” American Economic Review,
December 1985, 75(5), pp. 1056-70.
Canzoneri, Matthew B; Nolan, Charles and Yates, Anthony.
“Mechanisms for Achieving Monetary Stability: Inflation
Targeting versus the ERM.” Journal of Money, Credit, and
Banking, February 1997, 29(1), pp. 46-60.
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.

Friedman, Benjamin M. “The Use and Meaning of Words
in Central Banking: Inflation Targeting, Credibility, and
Transparency,” in P. Mizen, ed., Central Banking, Monetary
Theory, and Practice: Essays in Honor of Charles Goodhart.
Volume 1. Northampton, MA: Edward Elgar, 2003, pp.
111-24.
Friedman, Benjamin M. and Kuttner, Kenneth. “A Price
Target for U.S. Monetary Policy? Lessons from the
Experience with Money Growth Targets.” Brookings
Papers on Economic Activity, 1996, (1), pp. 77-125.
Friedman, Milton. “A Monetary and Fiscal Program for
Economic Stability.” American Economic Review, June
1948, 38(3), pp. 245-64.
Friedman, Milton. A Program for Monetary Stability. New
York: Fordham University Press, 1959.
Friedman, Milton “The Role of Monetary Policy.” American
Economic Review, March 1968, 58(1), pp. 1-17.
Gavin, William T. “Inflation Targeting: Why It Works and
How To Make It Work Better?” Business Economics, April
2004, 39(2), pp. 30-37.

Clouse, James; Henderson, Dale W.; Orphanides, Athanasios;
Small, David H. and Tinsley, P.A. “Monetary Policy When
the Nominal Short-Term Interest Rate Is Zero.” Topics in
Macroeconomics, 2003, 3(1), Article 12;
www.bepress.com/bejm/topics/vol3/iss1/art12.

Goodfriend, Marvin. “Monetary Mystique: Secrecy and
Central Banking.” Journal of Monetary Economics,
January 1986, 17(1), pp. 63-92.

Debelle, Guy. “The Australian Approach to Inflation
Targeting.” Working paper, Reserve Bank of Australia,
2003.

Goodfriend, Marvin and King, Robert G. “The Case for
Price Stability,” in A.G. Herrero et al., eds., The First
European Central Banking Conference: Why Price Stability?
Frankfurt: European Central Bank, 200l, pp. 53-94;
www.ecb.int/pub/pdf/whypricestability.pdf.

Erceg, Christopher J.; Henderson, Dale W. and Levin,
Andrew T. “Optimal Monetary Policy with Staggered Wage
and Price Contracts.” Journal of Monetary Economics,
October 2000, 46(2), pp. 281-313.
Faust, Jon. “Whom Can We Trust to Run the Fed?
Theoretical Support for the Founders’ Views.” Journal of
Monetary Economics, April 1996, 37(2), pp. 267-83.

Gramlich, Edward M. “Maintaining Price Stability.” Remarks
before the Economic Club of Toronto, Toronto, Canada,
October 1, 2003; www.federalreserve.gov/boarddocs/
speeches/2003/20031001/default.htm.
Greenspan, Alan. “Chairman’s Remarks.” Federal Reserve
Bank of St. Louis Review, July/August 2002, 84(4), pp. 5-6.

J U LY / A U G U S T 2 0 0 4

137

Faust and Henderson

Greenspan, Alan. “Opening Remarks” in Monetary Policy
and Uncertainty: Adapting to a Changing Economy,
Federal Reserve Bank of Kansas City Symposium
Proceedings, 2003.
Heikensten, Lars and Vredin, Anders. “The Art of Targeting
Inflation.” Swedish Riksbank Economic Review, Fourth
Quarter 2002, pp. 5-34.
Henderson, Dale W. “What Should Monetary Policy Be
Targeting,” in P. Minford, ed., Money Matters. Cheltenham:
Edward Elgar, 2004, pp. 312-22.
HM Treasury. Remit for the Monetary Policy Committee of
the Bank of England and the New Inflation Target.
Cheltenham: HM Treasury, December 2003.
Jensen, Henrik. “Targeting Nominal Income Growth or
Inflation?” American Economic Review, September 2002,
92(4), pp. 928-56.
Keynes, John M. A Tract on Monetary Reform. London:
Macmillan, 1923.
Kiley, Michael T. “Monetary Policy under Neoclassical and
New-Keynesian Phillips Curves, with an Application to
Price Level and Inflation Targeting.” Finance and Economics
Discussion Series No. 1998-27, Board of Governors of the
Federal Reserve System, 1998.
Kim, Jinill and Henderson, Dale W. “Inflation Targeting and
Nominal Income Growth Targeting: When and Why Are
They Suboptimal?” International Finance Discussion
Paper 719, Board of Governors of the Federal Reserve
System, 2002.
King, Robert G. and Wolman, Alexander L. “What Should
the Monetary Authority Do When Prices Are Sticky?” in
J.B. Taylor, ed., Monetary Policy Rules. Chicago: University
of Chicago Press, 1999, pp. 349-98.
Kohn, Donald L. “Comment on ‘Inflation Targeting in the
United States?’” in Ben S. Bernanke and Michael Woodford,
eds., The Inflation Targeting Debate. Chicago: University
of Chicago Press (forthcoming).
Kohn, Donald L. and Sack, Brian P. “Central Bank Talk: Does
It Matter and Why?” Presented at The Macroeconomics,
Monetary Policy, and Financial Stability Conference in
Honor of Charles Freedman, Bank of Canada, Ottowa,
Canada, June 20, 2003; www.federalreserve.gov/BoardDocs/
Speeches/2003/20030620/default.htm.
138

J U LY / A U G U S T 2 0 0 4

REVIEW
Kydland, Finn E. and Prescott, Edward C. “Rules Rather
than Discretion: The Inconsistency of Optimal Plans.”
Journal of Political Economy, June 1977, 85(3), pp. 473-91.
Leeper, Eric M. “An ‘Inflation Reports’ Report.” Sveriges
Riksbank Economic Review, Third Quarter 2003, pp.
94-118.
Levin, Andrew T. and Williams, John C. “Robust Monetary
Policy with Competing Reference Models.” Working Paper
2003-10, Federal Reserve Bank of San Francisco, 2003.
Lockwood, Ben; Miller, Marcus and Zhang, Lei . “Designing
Monetary Policy When Unemployment Persists.” Working
paper, University of Exeter, 1995.
Lucas, Robert E. Jr. “Econometric Policy Evaluation: A
Critique,” in K. Brunner and A.H. Meltzer, eds., The Phillips
Curve and Labor Markets. Volume 1 of the Carnegie
Rochester Conference Series on Public Policy. Amsterdam:
North Holland, 1976, pp. 19-46.
Lucas, Robert E. Jr. “Rules, Discretion and the Role of
Economic Advisers,” in S. Fischer, ed., Rational
Expectations and Economic Policy. Chicago: University of
Chicago, 1980, pp. 199-210.
McCallum, Bennett T. “Crucial Issues Concerning Central
Bank Independence.” Journal of Monetary Economics,
June 1997, 39(1), pp. 99-112.
McCallum, Bennett T. and Nelson, Edward. “Performance
of Operational Policy Rules in an Estimated Semiclassical
Structural Model,” in John Taylor, ed., Monetary Policy
Rules. Chicago: University of Chicago Press, 1999, pp.
349-98.
Melitz, Jacques. “Monetary Discipline, Germany and the
European Monetary System.” Kredit and Kapital, 1988,
21(4), pp. 481-511.
Meyer, Lawrence H. “Inflation Targets and Inflation
Targeting.” North American Journal of Economics and
Finance, August 2002, 13(2), pp. 147-62.
Mishkin, Frederic S. “International Experiences with
Different Monetary Policy Regimes.” Journal of Monetary
Economics, June 1999, 43(3), pp. 579-605.
Mishkin, Frederic S. “Commentary,” in Rethinking
Stabilization Policy, Federal Reserve Bank of Kansas City
Symposium Proceedings, 2002.

FEDERAL R ESERVE BANK OF ST. LOUIS

Obstfeld, Maurice. “Models of Currency Crises with SelfFulfilling Features,” European Economic Review, 1996,
40(3-5), pp. 1037-47.
Orphanides, Athanasios. “Monetary Policy Evaluation with
Noisy Information.” Journal of Monetary Economics,
April 2003a, 50(3) pp. 605-31.
Orphanides, Athanasios. “The Quest for Prosperity Without
Inflation.” Journal of Monetary Economics, April 2003b,
50(3), pp. 633-63.
Persson, Torsten and Tabellini, Guido. “Designing
Institutions for Monetary Stability.” Carnegie-Rochester
Conference Series on Public Policy, December 1993, 39,
pp. 53-84.
Persson, Torsten and Tabellini, Guido. Macroeconomic
Policy, Credibility and Politics. New York: Routledge, 2002.
Phelps, Edmund S. “Money-Wage Dynamics and LaborMarket Equilibrium.” Journal of Political Economy, August
1968, 76(4, Part 2), pp. 678-711.
Phelps, Edmund S. Inflation Policy and Unemployment
Theory. London: Macmillan, 1972.

Faust and Henderson

Implementing and Monitoring Inflation Targets.”
European Economic Review, June 1997a, 41(6), pp. 1111-46.
Svensson, Lars E.O. “Optimal Inflation Targets, ‘Conservative’
Central Banks, and Linear Inflation Contracts.” American
Economic Review, March 1997b, 87(1), pp. 98-114.
Svensson, Lars E.O. “Inflation Targeting As a Monetary
Policy Rule?” Journal of Monetary Economics, June 1999,
43(3), pp. 607-54.
Svensson, Lars E.O. “Independent Review of the Operation
of Monetary Policy.” Wellington, New Zealand: New
Zealand Treasury, Independent Review of the Operation
of Monetary Policy, 2001;
www.treasury.govt.nz/monpolreview/.
Svensson, Lars E.O. “What Is Wrong With Taylor Rules?
Using Judgment in Monetary Policy through Targeting
Rules.” Journal of Economic Literature, June 2003, 41(2),
pp. 426-77.
Svensson, Lars E.O. and Woodford, Michael. “Implementing
Optimal Policy through Inflation Forecast Targeting,” in
Ben S. Bernanke and Michael Woodford, eds., The
Inflation Targeting Debate. Chicago: University of Chicago
Press (forthcoming).

Rogoff, Kenneth. “The Optimal Degree of Commitment to
an Intermediate Monetary Target.” Quarterly Journal of
Economics, November 1985, 100(4), pp. 1169-89.

Sveriges Riksbank. Inflation Report 2003:4. Stockholm:
Sveriges Riksbank, 2003.

Romer, Christina D. and Romer, David H. “Federal Reserve
Information and the Behavior of Interest Rates.” American
Economic Review, June 2000, 90(3), pp. 429-57.

Truman, Edwin M. Inflation Targeting in the World
Economy. Washington, DC: Institute for International
Economics, 2003.

Rotemberg, Julio J. and Woodford, Michael. “An
Optimization-Based Econometric Framework for the
Evaluation of Monetary Policy,” in NBER Macroeconomics
Annual 1997. Cambridge, MA: MIT Press, 1997, pp. 297346.

Walsh, Carl E. “Optimal Contracts for Independent Central
Bankers.” American Economic Review, March 1995, 85(1),
pp. 150-67.

Steagall, Henry B. Remarks from The U.S. Congressional
Record. Washington, DC: U.S. Government Printing Office,
1935.
Stein, Jeremy C. “Cheap Talk and the Fed: A Theory of
Imprecise Policy Announcements.” American Economic
Review, March 1989, 79(1), pp. 32-42.
Svensson, Lars E.O. “Inflation Forecast Targeting:

Warburg, Paul M. The Federal Reserve System. Volume 2.
New York: Macmillan, 1930.
Welteke, Ernst. “Reforming Europe.” Closing remarks at
the European Banking Congress, Frankfurt, November 21,
2003; www.bis.org/review/r031202b.pdf.
Wolman, Alexander L. “A Primer on Optimal Monetary
Policy with Staggered Price-Setting.” Federal Reserve
Bank of Richmond Economic Quarterly, Fall 2001, 87(4),
pp. 27-52.

J U LY / A U G U S T 2 0 0 4

139

REVIEW

Faust and Henderson

Appendix

BACKWARD-LOOKING VERSION
In the backward-looking version, φ=1 and κ=0. Therefore, the loss function with the output gap
eliminated using the Phillips curve is
L tBL =
(A.1)

(

ltBL+ j = π t + j − π ∗

)

2

+

1  ` j BL 
εt  ∑ β lt + j 
2  j =0

[(

]

2
λ
−
−
−
−
.
π
π
β
π
π
ε
1
t
j
t
j
t
j
+
+
−
+
α2

) (

)

The first-order condition for π t+j is

πt+ j − π ∗ +
(A.2)

{[(π − π ) − (π − π )] − ε }
λβ
−
{[(π − π ) − (π − π )]} = 0.
α

λ
α2

t+ j

t + j −1

t+ j

t + j +1|t + j

2

t+ j

Taking the unconditional expectation of equation (A.2) yields the result that π–=π *. Using this result,
collecting terms, and multiplying through by –α 2 yields

βπ˜ t + j +1|t + j −
(A.3)

1
Aπ˜ t + j + π˜ t + j −1 = −ε t + j ,
λ

j = 0,1,...

A = α 2 + (1 + β )λ .

π˜ t + j = π t + j − π ∗,

The roots represented by Λ and Ψ are
˜− A
˜ 2 − 1 < 1,
0<Λ = A
β
(A.4)

˜ = 1 A,
A
2βλ

˜+ A
˜2 − 1 ,
1< Ψ = A
β

˜2 − 1 >0
A
β

lim Λ = 0,
λ →0

lim Λ = 1 .

λ →`

Using standard methods we obtain the solutions:
`

π˜ t + j = Λπ˜ t + j −1 + Λε t + j + Λ2 ∑ (βΛ )
i =0

(A.5)
˜yt + j =

Λ −1
Λ −1
π˜ t + j −1 −
εt + j ,
α
α

i

εt + j εt + j +1+i ,

j = 0,1,...

j = 0,1,...

where ỹt+j=yt+j – y P. A positive cost shock increases inflation and reduces output.
Taking expectations conditioned on information at time t yields the forecasts

π˜ t = Λπ˜ t −1 + Λε t ,
(A.6)

π˜ t + j +1|t = Λπ˜ t + j|t ,

Λ −1
Λ −1
π˜ t −1 −
εt
α
α
Λ −1
=
π˜ t + j|t ,
j = 0,1,...
α

˜yt =
˜yt + j +1|t

as stated in the text. By repeated substitution

π t + j|t = π ∗ + ( Λ )
(A.7)
yt + j|t = y P +

140

J U LY / A U G U S T 2 0 0 4

j +1

π ∗ + (Λ)

j +1

(π t −1 + εt ),

j = 0,1,...

Λ −1
Λ −1
Λ −1 
π t −1 −
εt 
(Λ) j π ∗ + (Λ) j 
α
α

 α

j = 0,1,...

FEDERAL R ESERVE BANK OF ST. LOUIS

Faust and Henderson

Now we derive the unconditional efficient policy frontier implied by the optimal commitment rule. We
begin by finding the unconditional variances of π̃ t+j and ỹ:

(Λ − 1) Λ2 + (Λ − 1) = 1 − Λ ,
Λ2
2
σ
=
,
˜y
α 2 1 − Λ2
α2
α 2 (1 + Λ )
1 − Λ2
2

(A.8)

σ π̃2 =

2

where σ ε2 has been set equal to one for simplicity. To complete the derivation, we invert the expression for
σπ̃2 to obtain an expression for Λ in terms of σπ̃2 and substitute that expression into the expression for σ ỹ2:

σ 2˜y =

(A.9)

σ π2˜ + 1 − Y
α

2

(

σ π2˜

+1+ Y

)

> 0,

((σ + 1)σ )

Y=

2
π˜

2
π˜

The policy frontier has a negative slope and is convex to the origin since
dσ 2˜y

(A.10)

dσ π2˜

=−

(

σ π2˜ + 1

α 2 Y σ π2˜ + 1 + Y

)

2

<0

( ) = 1  2σ + 1 + 2Y
2 α σ Y σ + 1 + Y
d (σ )
(
)


d 2 σ 2˜y

(A.11)


 > 0.
2


2
π˜

2 2
π˜

2

2
π˜

2
π˜

An increase in σπ̃2 is achieved through an increase in Λ (generated by an increase in λ), which lowers σ ỹ2.

FORWARD-LOOKING VERSION
In the forward-looking version, φ=0 and κ >0. The loss function with the output gap eliminated using the
Phillips curve is
L FL
t =
(A.12)

1  ` j FL 
εt  ∑ β lt + j 
2  j =0

(

ltFL+ j = π t + j − π ∗

)

2

+

[(

]

2
λ
π t + j − π − β π t + j −1 − π − ε t + j − ακ .
2
α

) (

)

To simplify the analysis, we assume that the only shock occurs in period t.

DISCRETION
Under discretion, the first-order condition for π t+j holding π t+j+1 constant is
(A.13)

πt + j − π ∗+

[(

]

λ
π t + j − π − β π t + j +1 − π − ε t + j − ακ = 0 .
α2

) (

)

Taking the unconditional expectation yields the familiar inflation bias result
(A.14)

π = π ∗+

λκ
α

so that the first-order condition for deviations from the unconditional mean is

(A.15)

β
1
1
1
πˆ t + j + λ  πˆ t + j − πˆ t + j +1 − ε t + j  = 0
α
α
α
α
πˆ t + j = π t + j − π ,
πˆ ∗ = π ∗ − π .

J U LY / A U G U S T 2 0 0 4

141

REVIEW

Faust and Henderson

It will be optimal to have no deviations from period t+1 on because there are no shocks then. Therefore
the solution for the deviation in period t is

π̂ t =

(A.16)

λ
εt
α2 + λ

and the full solutions for inflation in period t and all other periods are

πt+ j = π ∗ +

(A.17)

λκ
λ
+
εt ,
α α2 + λ

j =0,1,K

COMMITMENT
Under commitment, the first-order condition for any period except period t is
−

[(

] (

λβ
π t + j − π − β π t + j +1 − π − ε t + j − ακ + β π t + j +1 − π ∗
α2

(A.18)
+

) (

[(

)

)

]

λβ
π t + j +1 − π − β π t + j +2 − π − ε t + j +1 − ακ = 0 .
α2

) (

)

Taking unconditional expectations yields the familiar commitment result that

π = π ∗.

(A.19)

Multiplying through by − αλβ and collecting terms, the first-order conditions can be rewritten in deviation
form as
1
(A.20)
β πo t + 2 − A πo t +1 + πo t = ε t
λ
2

β πo t + j +2 −

(A.21)

1 o
A π t + j +1 + πo t + j = 0,
λ

j = 1, 2,K

o
∗
π t+ j = πt+ j − π ,

(A.22)

where the condition (A.20) for π8t+1 is the only one with a shock term. Note that the distortion term does
not enter these conditions. Solving the difference equation (A.21) yields
o

o

π t + j +1 = Λ π t + j ,

(A.23)

j = 1, 2,...

The first-order condition for period t in deviation form is
λ o

o
o
π t + 2  π t − β π t +1 − ε t − ακ  = 0.
(A.24)
α
Multiplying through by α 2, collecting terms, and rearranging yields

(

)

− λβ πo t +1 + α 2 + λ πo t = λκα + λε t .

(A.25)

The first-order conditions for π8t and π8t+1 can now be written as

(α

(A.26)

)

+ λ πo t − λβ πo t +1 = λκα + λε t
o
πt −

(A.27)

142

2

J U LY / A U G U S T 2 0 0 4

1
( A − λβΛ) πo t +1 = εt ,
λ

FEDERAL R ESERVE BANK OF ST. LOUIS

Faust and Henderson

where π8t+2 has been eliminated from equation (A.20) using equation (A.23) for j=1 and A and Λ are
defined in equations (A.3) and (A.4), respectively.
The solutions are
o
κ λκ
(A.28)
+ Cπε ,t ε t
π t = Cπ ,t
α
(A.29)

(A.30)

(A.31)

(A.32)

λκ

o
j κ
− Cπε ,t +1ε t  ,
π t + j +1 = Λ  Cπ ,t +1


α
Cπκ ,t =

α 2 ( A − λβΛ )
,
λ∆
Cπκ ,t +1 =

α2
,
∆

Cπε ,t =

j = 0,1,...

A − λβΛ − λβ
∆

Cπε ,t +1 =

α2
∆

 α2 
∆=
+ 1 C − λβ > 0, C = A − λβΛ > 0, A − λβΛ − λβ > 0
 λ


where all of the C’s are positive and the C’s for period t differ from those for period t+1 and all other periods.
First suppose that target output is above potential (κ >0) but that there is no shock (εt=0). It can be shown
that the inflation rate starts out above π* by less than under discretion and approaches π* asymptotically
and that output starts above y P and approaches y P asymptotically. Now suppose that κ =0 and εt >0. It
can be shown that inflation is above π* in period t, below π* from period t+1 on, and approaches π*
asymptotically.

J U LY / A U G U S T 2 0 0 4

143

REVIEW

144

J U LY / A U G U S T 2 0 0 4

Commentary
Benjamin M. Friedman

M

acaulay—not Frederick Macaulay, who
did economic research on interest rates,
but the great historian of the British
empire, Thomas Babington Macaulay—wrote that
the benefactors of mankind are customarily attacked
by “the dunces of their own generation” for going
too far, as well as by “the dunces of a future generation” for not going far enough.1 The consensus on
display at today’s conference, and perhaps more
broadly in the economics profession as well, is that
inflation targeting where it is already in practice
is, and wherever it is adopted in the future will be,
a significant benefactor of, if not mankind, then at
least monetary policy. Compared with that apparent
consensus, and with today’s paper by Jon Faust and
Dale Henderson, my view is that of Macaulay’s
dunces both of the present and of the future: I will
argue that inflation targeting goes too far—or, what
in this context amounts to the same thing, takes us
in a direction we should not want to go. And I will
also argue that while Faust and Henderson’s paper
contains many valid and important criticisms of
inflation targeting, they do not go nearly far enough
in following the logical implications of the criticisms
they offer.
Whether inflation targeting has led to superior
outcomes for monetary policy in countries whose
central banks have already adopted this practice is,
of course, an empirical matter.2 The paper presented
at this conference by Andrew Levin, Fabio Natalucci,
and Jeremy Piger (2004, p. 75) concludes that “inflation targeting (IT) has played a role in anchoring
inflation expectations and in reducing the intrinsic
persistence of inflation.” But there is also plenty of
1

Cited by Clive (1973, p. 481).

2

Faust and Henderson opened the version of their paper that they
presented at the conference by flatly declaring, “The inflation targeting
framework (ITF) has been a great success around the world.” In this
version they have abandoned that claim.

conflicting evidence. The recent paper by Laurence
Ball and Niamh Sheridan, for example, offers a quite
different interpretation of the same experience that
Levin et al. study: “This paper asks whether inflation
targeting improves economic performance, as measured by the behavior of inflation, output, and interest
rates...Once one controls for regression to the mean,
there is no evidence that inflation targeting improves
performance”.3
The main issues in Faust and Henderson’s
paper, however, are conceptual. In particular, they
continually—and rightly—highlight the role of
inflation targeting as a way for the central bank to
communicate with the public. They approvingly
quote Bernanke, Laubach, Mishkin, and Posen to the
effect that among the “important features of inflation targeting are vigorous efforts to communicate
with the public about the plans and objectives of the
monetary authorities...” (p. 118, emphasis added
by Faust and Henderson).4 They begin their own
paper by saying, “The core requirements of inflation
targeting are an explicit long-run inflation goal and
a strong commitment to transparency” (my emphasis). And they go on to say, “Not only are ITF [inflation targeting framework] central banks among the
most transparent in the world, they have experimented aggressively with ways to make communication with the public more effective” (p. 117).
I disagree. As typically practiced today, inflation
targeting is a framework not for communicating
the central bank’s goals and policies but for obscuring them. In crucial ways it is not a window but a
screen. It promotes not transparency—at least not in
the dictionary sense of the word—but opaqueness.
The key issue here, as Faust and Henderson
clearly understand, is multiple goals. Monetary policy
has one instrument: typically today some short-term
3

Ball and Sheridan (2003, abstract).

4

Bernanke et al. (1999).

Benjamin M. Friedman is the William Joseph Maier Professor of Political Economy, Harvard University.
Federal Reserve Bank of St. Louis Review, July/August 2004, 86(4), pp. 145-49.
© 2004, The Federal Reserve Bank of St. Louis.

J U LY / A U G U S T 2 0 0 4

145

REVIEW

Friedman

interest rate, but alternatively the quantitative change
in the central bank’s liabilities. As Tinbergen showed
decades ago, in the absence of degeneracy or other
pathologies, the solution to a problem with one
instrument and multiple targets can always be
expressed in terms of the intended trajectory for any
one arbitrarily chosen target. So far, so good. But
the question Tinbergen did not address is whether
that way of describing the solution promotes or
subverts public understanding of what the policymaker is doing, and why.
Faust and Henderson’s way of putting this
matter—which I like very much—is to think in terms
of the mean inflation rate and the variability of inflation. Inflation targeting communicates well about
mean inflation. As they point out, however, “agreement regarding the mean inflation rate has very
few practical consequences at any finite horizon”
(p. 117). By contrast, inflation targeting does not
communicate at all well about how much inflation
should vary, or why.
There are at least two reasons why policymakers
should expect, indeed want, inflation to vary. One,
of course, is the unpleasant fact of technical errors.
More central to this entire line of argument is the
policymaker’s concern for other goals of monetary
policy. (Within the literature of inflation targeting
this issue is made most explicit in Lars Svensson’s
formulation, in which the key decision is how rapidly
to bring inflation back to the desired rate after some
departure from it.5) The failure of most inflation
targeting schemes, as implemented by actual central
banks, to say anything about how much inflation
variability the central bank will tolerate, or why, is
also a failure to say anything about any goals of
monetary policy other than inflation, or about the
relationship between those goals and the inflation
goal.
Moreover, as I have argued elsewhere, I believe
this failure is intentional on the part of the central
banks that adopt this framework.6 As Faust and
Henderson put it, “One of the most famous principles
of strategic skewing in the folk wisdom of central
banking is that central banks should ‘do what they
do, but only talk about inflation’.” They go on to say
that “the ITF might be viewed as an application of
this folk wisdom. Without the folk wisdom, it is difficult to imagine why a policy of optimization with
5

Svensson (1997).

6

See Friedman (2003).

146

J U LY / A U G U S T 2 0 0 4

multiple conflicting goals would be called ‘inflation
targeting.’ Calling reports on all aspects of policy
‘inflation reports’ is an analogous misnomer. The
folk wisdom would also justify discussing goals
other than inflation only as they affect the horizon
over which one intends to hit the inflation target”
(p. 132).
They obviously have in mind, for example, the
Bank of England. The Bank of England, however, is
by no means the only central bank to exhibit this
form of anti-transparency. For example, the Bank of
Canada’s one-page public explanation of its policymaking framework, entitled “Canada’s InflationControl Strategy” and prominently printed on the
inside front cover of the Bank’s regular Monetary
Policy Report, has only three sentences bearing on
the strategy’s underlying rationale: “Inflation control is not an end to itself; it is the means whereby
monetary policy contributes to solid economic
performance. Low inflation allows the economy to
function more effectively. This contributes to better
economic growth over time and works to moderate
cyclical fluctuation in output and employment.”7
There is no mention of any tension, at any horizon,
between the Bank’s inflation goal and output,
employment, or any other matter of potential concern to monetary policy. (The remainder of the statement, devoted to operational considerations, also
gives no hint of any reason, beyond technical errors,
for inflation ever to depart from the desired rate.)
What is the import of all this? As Faust and
Henderson write, “the primary shortcoming of the
ITF communication policy is in making clear the
roles and balance of multiple goals...[T]he ITF as
implemented often involves elements that are literally inconsistent with best-practice policy and, in
any case, obfuscates some basic issues” (p. 124).
Further, “any discussion of stability of prices or inflation must inevitably raise issues of other goals...
[S]everal aspects of the ITF as practiced do not provide a natural and straightforward framework for
communicating this fact” (p. 126).
Given this assessment—which I believe is
correct—two questions follow: Is this aspect of
inflation targeting, as actually practiced, incidental
or deliberate? And, in the end, is it only about how
central banks talk—although that too is clearly
important—or does it also have implications for what
central banks do?
7

Bank of Canada, Monetary Policy Report, April 2003, inside front
cover page.

FEDERAL R ESERVE BANK OF ST. LOUIS

Faust and Henderson imply—and elsewhere I
have argued more directly—that the connection is
deliberate.8 As they note, inflation targeting appeared
on the policymaking scene at a time when the pressing need, throughout the industrialized world, was
to reduce the ongoing rate of inflation. As I have also
argued, the intellectual background against which
inflation targeting emerged consisted of the timeinconsistency discussion and the forward-looking
Phillips curve; and both of these lines of thought
naturally lend themselves to the kind of obfuscation
that inflation targeting embodies. The crucial implication of time inconsistency, not just for monetary
policy but for a broad class of problems (lender-oflast-resort policy, for example), is that misleading
people about the policymaker’s likely actions—if it
is possible to do so—can induce beneficial behavior.
But the same implication is also inherent in any
model based on the standard forward-looking Phillips
curve: The lower is the public’s expectation of future
inflation, the more favorable is the trade-off between
inflation and output that the policymaker faces in
the present—in other words, less inflation for given
output, or more output for given inflation.
Given the central role in macroeconomics now
played by the forward-looking Phillips curve, this
logic, in both simple and sophisticated forms, is pervasive. It is no surprise, for example, that in section
1 of Michael Woodford’s paper for this conference
(on “Advantages of an Explicit Target for Monetary
Policy”), section 1.1 is titled “Central Banking as the
Management of Expectations” (emphasis added).
This is not the place to go over yet again the concerns
I have expressed elsewhere about this way of thinking about monetary policy.9 The question to pose,
however, is whether, when the central bank in fact
has multiple goals but quantifies only one—indeed,
when it refuses to talk explicitly about any of the
others, except in terms of how they bear on the
achievement of that one—we should call this kind
of communications policy the management of
expectations or the manipulation of expectations.
As Woodford and many others have ably shown,
the public’s expectations matter for economic
behavior, including the efficacy of monetary policy,
and so even if all that the obfuscation inherent in
inflation targeting did were to affect expectations,
that in itself would be important. But there is also
8

See, again, Friedman (2003).

9

Interested readers can refer to my discussion of Eggertsson and
Woodford (2003).

Friedman

ground to believe that inflation targeting may distort
not just what the central bank says but what it does.
One reason for thinking so, which Faust and
Henderson note, is “the ITF premise that the threat
of public criticism affects the incentives of the central
bank and thereby the course of policy” (p. 129). But
as they also rightly point out, “The ITF communication policy is tilted heavily toward emphasis on stabilizing inflation” (p. 126). As a result, “Given the
skewing of communication in the ITF...a weak policymaker may find it safest to excessively smooth inflation.” Hence, “the ITF seems as likely to complicate
as to facilitate achieving a proper balance of multiple
goals by a weak policymaker” (p. 131).
I have little to add to this important (and, I believe,
correct) line of argument, other than to say that it
probably applies to strong policymakers as well as
weak ones—what Faust and Henderson call “policymaker aversion to criticism” is pervasive—and to
suggest that it calls back into question an often-made
claim that they dismiss out of hand at the outset of
their paper: namely, that in many contexts the debate
over inflation targeting is really a debate over what
properly belongs in the central bank’s preference
function. Faust and Henderson are at pains to distinguish what Mervyn King has colorfully called “inflation nutters” from what they here label “NETers.”10
For practical purposes, however, these two positions
are isomorphic. Their respective implications for
monetary policy are observationally equivalent.
There is also a second reason for thinking that
the inflation-targeting framework affects not just
what the central bank says but also what it does. Put
simply, the point is that language matters. David
Hume, who importantly influenced the shaping of
our discipline in its formative years, both directly
and even more so through his influence on Adam
Smith, had this to say about how skewed language
affected the central political issue in the Britain of
his day (monarchy versus republic): “The Tories have
been obliged for so long to talk in the republican stile
that they...have at length embraced the sentiments
as well as the language of their adversaries.”11
We are all familiar with instances in our own
day of the same phenomenon. For example, how
might research on monetary policy (and macroeconomics more generally) have evolved differently
if the particular assumption about expectations
10

See King (1997).

11

Hume (1741, p. 72).

J U LY / A U G U S T 2 0 0 4

147

REVIEW

Friedman

introduced by Muth and Lucas had been labeled
“super-smart-agents expectations,” or, perhaps more
even-handedly, “model-consistent expectations,”
rather than the far more compelling “rational expectations”? Might the work now exploring the implications of “bounded rationality” have developed
earlier, or differently, under a less biased label?
To return to the case at hand, it is not too great
a leap to conjecture that one consequence of constraining the discussion of monetary policy to be
carried out entirely in terms of an optimal inflation
trajectory will be that concern for real outcomes
will atrophy, or even disappear from policymakers’
consideration altogether. Nor is it unreasonable to
suppose that the hope that this eventuality will ensue
is, for some advocates, a motivation for favoring
inflation targeting in the first place.
I shall turn in closing to three narrower and
more specific comments on Faust and Henderson’s
paper. First, they write that, “as a profession we are
more certain about our advice regarding the mean
of inflation” (p. 118) than about what we say regarding the variance (in other words, economic stabilization). This may well be true. But even so, a reader
of the relevant theoretical and empirical literature
is entitled to ask just how confident we are on this
score. At the conceptual level, there are at least five
reasons for choosing a mean inflation rate different
from zero: (i) measurement bias; (ii) the “stabilization
buffer” argument that Michael Woodford and Gauti
Eggertsson have recently analyzed at length12; (iii)
the role of inflation as “grease to the labor market,”
as famously argued in James Tobin’s AEA presidential
address and more recently highlighted by Akerlof,
Dickens, and Perry13; (iv) the distortionary tax argument, to which Stephanie Schmitt-Grohé has already
referred in her comments on Michael Woodford’s
paper at this conference; and (v) the fact that in the
United States today the principal asset bearing a
permanently fixed nominal interest rate (zero) is
currency, together with the apparent facts that much
of the outstanding U.S. currency is held outside the
country and that much of the rest is used by drug
dealers and other criminals on whom we should
want to impose distortionary taxes. At the empirical
level, there is no evidence that mean inflation even
quite far above zero by the standards of today’s industrialized world retards economic growth; Robert
12

Eggertsson and Woodford (2003).

13

Tobin (1972); Akerlof et al. (2000).

148

J U LY / A U G U S T 2 0 0 4

Barro’s work on this question has shown no effect
on growth associated with mean inflation up to 15
percent per annum, and Michael Sarel’s work has
shown no effect up to 8 percent.14
Second, while Faust and Henderson are certainly
correct that a belief in “long and variable lags” is
nowadays part of the common ground of monetary
economics, the familiar attempt to appeal to this
argument as a rationale for inflation targeting is at
best out of place and, more likely, misleading. Milton
Friedman’s classic argument applied not merely to
the attempt to vary the policy instrument in order
to control output, but also to control inflation.
Nothing is lost, or even changed, by rewriting the
notation in Friedman’s 1953 paper to make the lefthand-side variable π and the key right-hand-side
variable either r or M. The force of the long-andvariable-lags argument is the implied optimality of
a constant instrument rule (most famously, a constant money growth rule). Long and variable lags do
not constitute an argument for inflation targeting.
Third, Faust and Henderson’s point about the
symmetry of costs applying not just to the mean of
inflation but also to the variability of inflation is both
interesting and important. Referring to the target
range that the central bank announces for inflation,
they rightly point out that “excessive frequency of
being inside the range is also evidence of misbehavior.” An appropriate analytical framework recognizes
that “excessive smoothness and excessive volatility
of inflation are equally costly at the margin in
equilibrium” (p. 125, emphasis added).
In conclusion, I disagree, sharply, with what
increasingly looks like an emerging consensus that
inflation targeting is, if not the optimal framework
for monetary policy, then a close enough approximation to be about as good a framework as any realworld central bank can practically hope to have.
More specifically, I do not believe that inflation targeting is a framework that the Federal Reserve System
should adopt for the United States. For the many
reasons I have explained in the course of discussing
Faust and Henderson’s paper, my view of inflation
targeting is that of Macaulay’s dunce of the present:
I think inflation targeting would take U.S. monetary
policy too far, in a direction in which we should not
want to go. And in regard to their paper, I am content
to be a dunce of the future. Faust and Henderson
have all the right insights. They fall short only in
not following the implications of these insights far
enough.
14

Barro (1995); Sarel (1996).

FEDERAL R ESERVE BANK OF ST. LOUIS

REFERENCES
Akerlof, George A.; Dickens, William T. and Perry, George L.
“Near-Rational Wage and Price Setting and the Long-Run
Phillips Curve.” Brookings Papers on Economic Activity,
2000, (1), pp. 1-44.
Ball, Laurence and Sheridan, Niamh. “Does Inflation
Targeting Matter?” NBER Working Paper No. 9577,
National Bureau of Economic Research, 2003.
Barro, Robert J. “Inflation and Economic Growth.” Bank of
England Quarterly Bulletin, May 1995, pp. 166-76.
Bernanke, Ben S.; Laubach, Thomas; Mishkin, Frederic S.
and Posen, Adam S. Inflation Targeting: Lessons from the
International Experience. Princeton: Princeton University
Press, 1999.
Clive, John. Macaulay: The Shaping of the Historian. New
York: Knopf, 1973.
Eggertsson, Gauti B. and Woodford, Michael. “The Zero
Bound on Interest Rates and Optimal Monetary Policy.”
Brookings Papers on Economic Activity, 2003, (1), pp.
139-211.

Friedman

Friedman, Milton. “The Effects of a Full-Employment Policy
on Economic Stability: A Formal Analysis,” in M. Friedman,
ed., Essays in Positive Economics. Chicago: University of
Chicago Press, 1953.
Hume, David. “Of the Parties in Great Britain,” in E.F. Miller,
ed., Essays: Moral, Political, and Literary. Indianapolis:
Liberty Fund, 1741 [reprinted 1985].
King, Mervyn. “Changes in UK Monetary Policy: Rules and
Discretion in Practice.” Journal of Monetary Economics,
June 1997, 39(1) pp. 81-97.
Levin, Andrew T.; Natalucci, Fabio M. 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.
Sarel, Michael. “Nonlinear Effects of Inflation on Economic
Growth.” IMF Staff Papers, March 1996, 43(1), pp. 199-215.
Svensson, Lars E.O. “Inflation Forecast Targeting:
Implementing and Monitoring Inflation Targets.” European
Economic Review, June 1997, 41(6), pp. 1111-46.
Tobin, James. “Inflation and Unemployment.” American
Economic Review, March 1972, 62(1/2), pp. 1-18.

Friedman, Benjamin M. “The Use and Meaning of Words
in Central Banking: Inflation Targeting, Credibility and
Transparency,” in P. Mizen ed., Central Banking, Monetary
Theory and Practice. Cheltenham: Edward Elgar, 2003.

J U LY / A U G U S T 2 0 0 4

149

REVIEW

150

J U LY / A U G U S T 2 0 0 4

Practical Problems and Obstacles to Inflation
Targeting
Laurence H. Meyer

T

he number of conferences, papers, and
speeches on inflation targeting suggests a
growing interest in exploring whether and
in what way the Federal Open Market Committee
(FOMC) should consider adopting an explicit numerical objective for inflation. Because the devil is often
in the details, it is important to go beyond the general interest in this direction and to explore obstacles
to moving in the direction and, at the same time,
to begin to think about some practical problems
that would have to be resolved if such an approach
were to be implemented.
My point of departure is the conviction that, if
the FOMC were to adopt an explicit numerical
inflation target, the vision of the resulting regime
would have to fit both the political realities and the
basic approach to monetary policymaking in the
United States over the past decade. Indeed, the case
for adopting an explicit inflation target in the United
States is typically rationalized in terms of continuity
rather than change. That is, it is an attempt to ensure
continuity in the conduct of monetary policy, especially after the departure of Alan Greenspan, not to
be an instrument for changing the way in which
monetary policy has been conducted over the past
decade or two.
The key distinction essential for understanding
the regime that would be a good fit for the United
States is between inflation targets and inflation targeting. After explaining that distinction, I will offer my
view of the vision of the Greenspan FOMC and consider the consistency of that vision with a regime
with an explicit numerical inflation target. Next I
consider the political climate for adopting an inflation target and other potential obstacles. I will conclude with a consideration of implementation details,
as the choice is not ultimately between an explicit
and implicit target in principle, but between the
current practice and a specific alternative.

INFLATION TARGETS AND INFLATION
TARGETING
This distinction between inflation targets and
inflation targeting, first made in a speech in July 2001
while I was a member of the Board of Governors,
can perhaps be best understood in terms of a two-bytwo matrix (Table 1). Across the top, I identify two
types of inflation targets, one implicit (like the
United States today) and the other explicit (as in socalled “inflation targeting” countries today). Down
the side I identify two forms of mandate that central
banks around the world operate under. These mandates are typically set by the legislatures. The United
States and Australia operate under a dual mandate,
according to which 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. Inflation-targeting countries generally
operate under hierarchical mandates, one in which
price stability is identified as the principal objective,
and central banks are restricted in pursuing other
objectives unless price stability has been achieved.
The United States has an implicit inflation target
and a dual mandate, the upper left. The United
Kingdom, Canada, and other so-called inflationtargeting countries have an explicit inflation target
and a hierarchical mandate. They are in the lower
right. Australia has a dual mandate along with an
explicit inflation target. That is the combination I
am suggesting for the United States.
Lars Svensson always responds to my proposal
by telling me that my distinction between dual and
hierarchical mandates is too strict. In particular it
misses the evolution in practice around the world.
In general, inflation-targeting countries today have
moved away from the initially austere implementation, more in line with the spirit of a hierarchical
mandate, and have become flexible inflation targeters, close cousins of dual mandate central banks.

Laurence H. Meyer is a distinguished scholar at the Center for Strategic and International Studies and president of Meyer’s Monetary Policy Insights.
Federal Reserve Bank of St. Louis Review, July/August 2004, 86(4), pp. 151-60.
© 2004, The Federal Reserve Bank of St. Louis.

J U LY / A U G U S T 2 0 0 4

151

REVIEW

Meyer

Table 1
Monetary Policy Regimes
Inflation target
Mandate

Implicit

Explicit

U.S.

Australia

Dual
Hierarchical

U.K., Canada

There are two possibilities in connection with
that evolution. First, the language in the mandates
was intended to impose restrictions on the central
banks, restrictions that went beyond simply identifying an explicit numerical inflation objective. So,
whatever you want to call inflation targeters, from
nutters to flexible inflation targeters, central banks
under a hierarchical mandate are going to behave
differently, specifically with less flexibility than in
the case of dual mandate central banks. It is of course
a topic for research, as to whether or not such differences can be identified in practice.
Second, what about the assessment that there
is no practical difference between flexible inflation
targeters operating under a hierarchical mandate
and dual mandate central banks? If this were the case,
the only difference between dual and hierarchical
mandate central banks is one of transparency. Dual
mandate central banks are transparent about their
objectives and flexible inflation targeters are not.
To allow for this possibility, I have shown a second
two-by-two matrix, what I call my Lars Svensson
model.
The transparency with respect to the stabilization objective for monetary policy is itself a fascinating subject. Historically, central bankers, unlike
members of the Congress, have been embarrassed
to admit they care about anything other than price
stability or conduct monetary policy for any other
purpose. Bob McTeer has perhaps said it best when
he reminded the Committee that “only hawks go to
central bank heaven.” Even in the United States, it
is much easier to find quotes from FOMC members
about the importance of price stability than about
the responsibility of monetary policymakers for
damping fluctuations in output around full employment. You all, I expect, recall the pillorying Alan
Blinder received when he noted at the Jackson Hole
conference in August 1994 that monetary policymakers should keep an eye on the unemployment
rate as well as on inflation.
152

J U LY / A U G U S T 2 0 0 4

This also reminds me of an incident at my very
first Jackson Hole conference as a member of the
Board in August 1996. Two of the leading central
bankers in the world took me aside to help educate
me about how to conduct myself so I would be
viewed as an upstanding central bank citizen. They
offered me the very same advice. Good central
bankers never admit they pursue stabilization policy.
Such an admission would reduce the confidence of
the public in your commitment to price stability and
therefore undermine your credibility and effectiveness as a monetary policymaker. I responded that I
appreciated the advice, especially from such distinguished central bankers, but that it left me a bit
confused. They seemed to be telling me that the
way to build credibility was to lie, specifically about
how I understood the objectives and how I intended
to conduct monetary policy. I never followed their
advice and indeed tried to educate the public about
the importance of the dual mandate.
This distinction between dual and hierarchical
mandates is central to the issue of the obstacles to
moving to an explicit numerical inflation target in the
United States and to the goal of designing a regime
that provides for continuity with the vision of monetary policy as practiced in the United States for at
least the past decade or two. Indeed, many of those
who vigorously oppose inflation targets do so because
they identify that practice with the hierarchical
mandates and a down-weighting of responsibility
of the central bank for promoting full employment.
One caveat is in order here. There are important
differences between the full employment and price
stability objectives, and I do not want to minimize
or disregard these differences, because they are
central to good practice for central banks. These
differences may indeed be the origin for hierarchical
mandates, though I expect the origin has more to do
with the disappointing experience with monetary
policy and inflation before the inflation-targeting
regimes were adopted.
First, a central bank, over some appropriate
intermediate term, can achieve an inflation target,
with a significant degree of precision. It has a choice
as to whether that target should be 2 percent or 3
percent or some other number. In a word, with
respect to inflation, the buck literally does stop at
the central bank. Central banks have less influence
over the short-run path of output and employment,
but, nevertheless, at the margin, can damp movements in output around its potential level.

FEDERAL R ESERVE BANK OF ST. LOUIS

Second, with respect to an inflation target, central
banks know where they want to go. Notwithstanding
biases and measurement issues, the central bank
can pick a target and get there. Unfortunately, the
same cannot be said for the full employment objective. We do not know exactly where it is at a given
moment or where it may be in the future. It is, of
course, not really as murky as that characterization
suggests, but we have only an estimate of the nonaccelerating inflation rate of unemployment (NAIRU)
and of potential output, and we have to update that
estimate over time, in part using information based
on the experience with inflation. This measurement
uncertainty does not mean that a central bank
should not pursue its estimate of full employment,
but it does imply that that pursuit has to be different
in some subtle ways from the way it pursues its inflation objective. In particular, central banks cannot
simply aim for a particular unemployment rate and
decide after the fact if it is really sustainable. Rather,
monetary policymakers have to be prepared to move
aggressively into the range of the estimate of full
employment and then perhaps move more gingerly
toward the estimate, watching each step along the
way for feedback as to whether it has gone far
enough or has overshot.

THE VISION OF GREENSPAN
My premise is that the goal of any change with
respect to the inflation target is one designed to
preserve and even ensure continuity in the way
monetary policy has been conducted under the
Greenspan FOMC. I therefore set out my vision of
that approach, identifying the three principles that,
in my view, have guided practice.
1. Build a reputation for a commitment to
price stability in order to anchor inflation
expectations. While an inflation target can

in principle contribute to this end, inflation
expectations, in practice, are based more on
performance than promise. Hence, outcomes
are more important than rhetoric. Therefore,
the first principle is that monetary policy
should be conducted to move the inflation rate
over time to a low, stable rate (the FOMC’s
implicit inflation target) and then to maintain
it close to that rate, with allowances for normal
cyclical variation. I thus identify the FOMC
under Greenspan as an implicit inflation
targeter.
2. Monetary policymakers should aggressively
respond to demand shocks that would

Meyer

otherwise move output and employment
from their full employment levels, with
appropriate consideration for prevailing
and prospective inflation rates. Well-

anchored inflation expectations provide
monetary policymakers increased freedom
to adjust policy in the shorter run to damp
movements in output relative to potential,
without concern that such aggressive use of
stabilization policy could destabilize inflation
expectations. There is a corollary to the
second principle. The anchoring of inflation
expectations itself makes the economy more
stable, reducing the effect on overall inflation
of adverse supply shocks and reducing the
instability that arises when the economy is
allowed to overheat and inflation rises above
the implicit inflation target, to be reversed
later on.
3. Monetary policymakers should be flexible
and pragmatic in the conduct of monetary
policy. Policy rules can provide useful guidance to monetary policymakers, but policymakers’ judgment will be essential in
responding to unique shocks or circumstances
and to making policy when the uncertainty
about the model, parameters, or the measurement of key variables becomes especially
large.
The Chairman, in my view, also believes that low,
stable inflation contributes to strong productivity
growth and hence to a higher maximum sustainable
rate of economic growth. This provides still another
reason why maintaining low stable inflation has
significant payoffs for economic performance. I
expect that the other members of the FOMC have
less faith in this principle than the Chairman.
What is unique about the Greenspan vision are
the synergies presumed between the two objectives
for monetary policy—price stability and damping
fluctuations around full employment—as well as
between price stability and achieving maximum
sustainable growth. What is also unique is that the
Chairman, based on this vision, is generally viewed
as being a hawk when it comes to containing inflation and a dove when it comes to quickly providing
support for a weakening economy. That is a remarkable combination, politically as well as economically,
and one that the FOMC presumably would not want
to lose as it considers adopting an explicit numerical
target for inflation.
J U LY / A U G U S T 2 0 0 4

153

REVIEW

Meyer

THE POLITICS OF INFLATION TARGETS

WHY BOTHER?

The Congress sets the objectives for monetary
policy, just as legislatures typically do in the case of
other central banks around the world. The Greenspan
vision, if not rhetoric, is, in my view, very much in
sync with the Congressional mandate.
There is, in my opinion, no chance that the
Congress would accept a regime with a hierarchical
mandate that raised the profile of price stability
and diminished the responsibility of the FOMC for
stabilization policy. It is true that there have been
specific bills introduced in the Congress that would
have moved the Fed in this direction. But those bills
reflected a minority position, indeed a very small
minority position, and the overwhelming majority of
the Congress would have rejected such an approach.
The only exception would be if there was a period
in which monetary policy in the United States was
not appropriately disciplined and inflation rose to
very high levels. The Congress might then impose a
more restrictive mandate. And that is the historical
experience that preceded the implementation of
many of the inflation-targeting regimes around the
world.
Nevertheless, there could be an obstacle in
achieving a consensus with the Congress about any
change in the conduct of monetary policy. The
greater danger is that the Congress would want to
balance an explicit target for inflation with an explicit
target for full employment. And, for the reasons
developed above—specifically, we do not know precisely the level of the NAIRU or potential output,
and our estimates of these indicators of full employment change over time—the FOMC could not accept
an explicit numerical target for inflation if it were
bundled with an explicit target for full employment.
This is perhaps the most important reason why
consultations with the Congress are so important
as a part of any interest of the FOMC in moving in
this direction. My belief is that the Congress would
accept an explicit numerical target for inflation in
the context of a reaffirmation of the Federal Reserve’s
responsibility for promoting full employment.
So both the political realities and the focus on
continuity require that an explicit numerical target
for inflation be implemented as part of a dual mandate and be done in a way that does not undermine
the flexibility of monetary policy to respond to various shocks or unusual circumstances. This is both
the only choice and the best direction.

The FOMC is already an implicit inflation targeter; and policy has been successful in achieving
a low rate of inflation, while preserving flexibility
to pursue stabilization policy. As a result, there is
some understandable skepticism about the payoff
in terms of better policy or improved outcomes
from making the inflation target explicit.
First, at the margin, an explicit inflation target
should contribute to anchoring inflation expectations, both by identifying the point at which the
public should put down the anchor and by establishing a consensus on the Committee about where
the anchor should be. I personally believe that an
explicit inflation target is more effective in anchoring
inflation expectations once the target has already
been achieved, rather than in lowering the cost of
initially achieving the target.
Second, an explicit inflation target would also
ensure a consensus on the Committee about the
inflation rate members should be aiming at, ensuring
that everyone is pushing in the same direction with
respect to the inflation objective. This could, at the
margin, improve the coherence of the deliberations
and the policy outcomes. It should be noted, however, that the payoff from more coherent internal
deliberations could be achieved by having an internally acknowledged target and does not require that
the target be made public. However, in my view, it
would be difficult to sustain and inappropriate politically for the Committee to agree on a target internally
and not announce it publicly.
Third, the added transparency about monetary
policy might further enhance the ability of bond
market participants to anticipate the future course
of monetary policy, shortening the lags from policy
to outcomes, and thereby improving the effectiveness
of policy. There is, I believe, a synergy between
transparency and policy effectiveness; and, if so, the
adoption of an explicit numerical inflation target
will be a step toward a more effective monetary
policy, improving the partnering between monetary
policy and the bond market.

154

J U LY / A U G U S T 2 0 0 4

PROCESS
Part of the successful navigation through the
political process is to set up a well-structured process
of inside deliberation, outside consultation, and
Congressional oversight about both the general
direction and the details of the implementation. The
degree of acceptance of the direction will ultimately

FEDERAL R ESERVE BANK OF ST. LOUIS

depend importantly on the details, including the
reaffirmation of the commitment to the dual mandate, the precise level of the target, whether there
is a range or just a point, the timeframe over which
the FOMC would be judged about its compliance
with the target, and any reporting requirements
that accompanied the new regime.
Below, I set out a possible process for moving
toward an explicit numerical inflation target in the
United States.
1. The process should begin with internal discussion. There has to be, at the outset, sufficient
support within the FOMC to justify a renewed
and intensified focus on the topic. The FOMC
then has to direct the staff to develop options
and implementation details.
2. The staff should then revisit some of the
topics they have previously considered,
update some of the previous relevant studies,
including the work presented at the July 1996
and July 1997 FOMC meetings, and set out
options for implementation details.
3. The FOMC should then make a decision as
to whether they want to adopt an explicit
inflation target and set out a preliminary
proposal for implementation details.
4. At this point, the Chairman should brief
Congressional leaders about the desire of
the FOMC to move in this direction and set
up a mutually agreed consultation process,
including hearings. An agreement would be
reached as to whether the inflation target was
to be adopted by the FOMC, in pursuance of
the existing Congressional mandate for price
stability, or whether the target was subject to
consideration and approval by the Congress.
5. The preliminary proposal should be released
to the public for comments.
6. After comments, a final proposal would be
released and, after further Congressional
consultations and perhaps hearings, would
be implemented.

OBSTACLES
The most obvious obstacle to establishing an
explicit numerical target for inflation is, of course,
Alan Greenspan. He has made it clear that he is
opposed to moving in this direction, though the
argument he made at a conference at this Bank—
specifically that he opposed an explicit target for
inflation because we could not measure inflation

Meyer

precisely enough—was singularly disappointing
and uncompelling. No matter. The Chairman clearly
prefers the status quo for the remainder of his term,
and no one on the Committee, including myself
when I was there, would push to adopt an explicit
inflation target while he was at the helm. But when
the Chairman’s term is over in early 2006, the topic
will likely resurface and become an active one inside
and outside the FOMC.
The second obstacle could be the new chairman.
The new chairman should presumably be given
some time to develop his own views on the topic
and will undoubtedly have a considerable influence
on whether the Committee moves in this direction.
On the other hand, I expect the Committee will be
looking to assert greater influence on policy outcomes and directions for policy strategy, and the
momentum inside the Committee to at least give
this careful consideration is likely to be impossible
to contain.
The third obstacle is the politics of inflation
targeting. The irony is that it might take a chairman
with the clout and political savvy of Alan Greenspan
to navigate such a change through the political
process. I believe that the current legislative mandate
provides a legal basis for the Fed to adopt a numerical inflation target, as long as the FOMC continues
to accept the dual mandate. Nevertheless, adopting
an inflation target would be viewed as an important
change in the monetary policy regime and, as such,
would need to be vetted with the oversight committees in the Congress. While I do not believe that
new legislation is needed, the Fed would have to
ensure that the Congress was comfortable with this
direction.
The fourth is inertia. Members of the FOMC
undoubtedly believe, as I do, that the Committee
has conducted policy in a flexible yet disciplined
and effective manner over the past decade. There
is no perceived imperative to change the policy
regime. It could be argued that adopting a numerical inflation target is not fundamentally a change
in the regime, but the point is still “if it ain’t broke,
why fix it.”
The fifth obstacle is the challenge of building a
consensus for the change inside the FOMC and then
for the details of the change. To do so, it will be
necessary to meet head on the legitimate concerns
of some who have staked out positions against such
a direction. By the way, my basic procedural proposal is to lock Governors Ben Bernanke and Don
Kohn in a room and not let them out until they have
J U LY / A U G U S T 2 0 0 4

155

REVIEW

Meyer

reached an agreement. That agreement is one I am
sure I and the overwhelming majority of the Committee likely could accept.
What is the core of the case against an inflation
target? Don Kohn, in my view, has presented the
most thoughtful argument against moving in this
direction. There may be a trade-off between becoming more transparent and accountable by adopting
an explicit numerical inflation target and losing
some of the flexibility that the Committee has had
in the conduct of monetary policy.
My proposal—adopting an explicit numerical
inflation target in the context of a reaffirmation of
the Committee’s commitment to the dual mandate—
is designed to meet that concern by making clear
that the intention of the change was not to alter the
way in which monetary policy has been conducted,
but only to make that conduct more transparent
and accountable.
Still, that concern lingers. It can perhaps be
appreciated in terms of the Taylor rule, viewed as a
simple summary of the way in which the FOMC
conducts monetary policy. The question then is
whether the FOMC can make explicit the numerical
target for the inflation objective—one of the key
terms on the Taylor rule—without, at the same time,
also altering the response coefficient on the output
gap relative to the response coefficient on the gap
between inflation and the inflation target. That is, can
the Committee more precisely identify one target
without changing the way it balances its two objectives and the aggressiveness, in particular, with which
it responds to deviations in inflation from its target?
Perhaps even more to the point, does adopting
an explicit and numerical inflation target force
monetary policymakers to be more mechanical in
their conduct of monetary policy, as in following
more closely a Taylor rule, as opposed to having
the flexibility to deviate from the rule when circumstances encourage the Committee to do so?
I would not argue that this is a trivial question
and one without merit. Indeed, many who favor an
inflation target or a full-fledged inflation-targeting
regime do so precisely because such an approach
constrains discretion. It is noteworthy that in their
discussions of the policy framework, Governor
Bernanke’s highest praise goes to one which involves
“constrained discretion,” while Governor Kohn
reserves his highest praise for a policy that is flexible
and pragmatic. Of course, they both undoubtedly
see the merit in the attempt to achieve a balance
among these properties of a policy regime.
156

J U LY / A U G U S T 2 0 0 4

I do not believe that, under my proposal, there
would be much risk that monetary policy would lose
its current flexibility—but that would depend on how
the change was understood by the Committee, the
Congress, and the public. My recent experience
reinforces this point. I have recently talked to economists who often say that they oppose moving to
an inflation-targeting regime; when they heard my
proposal, however, they not only indicated that
they could support it but seemed at least modestly
enthusiastic about going in that direction. That
suggests that much of the opposition to an explicit
numerical inflation target is really opposition to
the hierarchical mandates and perceived practices
of so-called inflation-targeting regimes.
In any case, I believe that the Committee would
have to become comfortable that they could conduct
policy with the degree of flexibility they have in
recent years, while still adopting an explicit numerical inflation target.
The last obstacle to adopting an inflation target
is agreeing upon the details. As it is often said, the
devil is in the details. Even those who might support
some version of an inflation target might not be
able to agree on the details of such a regime. That
provides a good bridge to my last section, practical
problems with implementing an inflation target.

PRACTICAL CONSIDERATIONS
I presume that the staff would be asked to come
up with some recommendations and perhaps
options for the various implementation details
required to develop a proposal for the adoption of
an explicit numerical inflation target. I will offer
my own views on how the various issues might be
resolved, identifying some potential deal busters,
but I would quite likely change my mind about some
of the details after reading the staff’s recommendations and hearing the comments both from members
of the Committee and outsiders who have been
focused on this topic.

What Price Index Should the Inflation
Target Be Based Upon?
I do not believe there is a definitive answer to
this question, but I also do not believe that the
answer is very important, assuming the choice is
between a broad production-based index, such as
the chain-weighted price index for gross domestic
product, or a broad-based consumption measure,
such as the consumer price index (CPI) or the personal consumption expenditures (PCE).

FEDERAL R ESERVE BANK OF ST. LOUIS

I do not believe that economic theory establishes
whether a production- or a consumption-based
measure of inflation is better as a target. Empirical
analysis might reveal interesting differences in the
way that monetary policy would respond to shocks
under production- and consumption-based measures, and that analysis might help to make the decision. For example, the response of monetary policy
to changes in the price of oil would be more aggressive under a consumption-based measure, although
that conclusion would be reversed if the target was
expressed in terms of a core measure of consumer
price inflation.
Still, I expect, as with all other countries that have
an inflation target, the choice will be a consumptionbased measure, as these appear more widely understood by the public. This is also the direction of the
discussion of this topic at FOMC meetings, specifically in July 1996.
This would leave us with a choice between the
CPI and PCE measures. I viewed this as a close call
up until the release of the chain version of the CPI.
The chain CPI inflation rate lined up much closer
to, and indeed very close to, the PCE measure. I
would therefore opt for the PCE measure. But I
wouldn’t be a fanatic about this choice. There are
times when I might, were I a member of the FOMC,
indicate that the Committee was putting somewhat
more attention on the CPI measure, as a result of
distortions believed to be affecting the PCE relative
to the CPI.

Should the Target Be Defined as
Applying Over a Specified Time Horizon?
The Congress and the public, as well as the
FOMC itself, are going to want to monitor the success
of the FOMC in achieving the target established for
inflation.
First, the Committee should always refer to inflation in terms of the 12-month inflation rate for the
measure it selects, and specifically not talk about
monthly or even quarterly inflation rates. All monitoring of FOMC performance relative to the target
should be focused on the 12-month rate.
Second, the Committee should emphasize that
it is focused on achieving the target over an intermediate term and will move only gradually to return
inflation to the target if a shock pushes inflation
away from the target, especially if it is pushed outside its monitoring range.
Many inflation-targeting countries explicitly
interpret their inflation target as applying to the

Meyer

intermediate term, typically out 11/2 to 2 years. This
is sometimes referred to as inflation-forecast targeting. Central banks often report their inflation forecast over this horizon, and it is expected that such
forecasts will be lined up on the inflation target.
This approach creates a potential tension with
a dual mandate. Under such a regime, it is not appropriate to always be at the inflation target, just as it
would not be appropriate to always be at the point
of maximum sustainable employment. An example
would be that it would make sense for inflation to
be above the inflation target late in the expansion.
If inflation just rose to the target during expansions
and fell below the target in recessions, the average
inflation rate will be below the target. In addition,
the overall target might have to be set higher in this
case, to reduce the prospect of occasionally hitting
the zero nominal bound.
In my view, a good way to set the target would
be as an average over the business cycle. Of course,
taken literally, that would be an average inflation
target, with properties like a price level target, in
that past deviations from the inflation target would
not be forgiven, but might at least implicitly be
expected to be offset by deviations in the other
direction later.

Should the Target Be the Overall
Measure of Inflation or a Core
Measure of Inflation?
If the objective is viewed as the forecast for
inflation over the intermediate term, say 11/2 to 2
years out, then it does not matter very much, if at
all, whether the target is specified as overall or core
inflation. That is because any shock will have dissipated by then, so the policy that would be consistent
with achieved overall and core inflation rates 2
years out would be very close, if not identical.
Still, the public and the Committee are going to
want to monitor inflation outcomes along the way
to determine whether the inflation performance is
broadly consistent with the target. In my judgment,
the core measure, by providing the best guidance
about expectations for overall and core inflation in
the future, is the better measure for monitoring how
the FOMC is doing relative to its inflation target. I
would prefer to use a core measure as the inflation
target itself because I believe that would reduce
possible confusion about how the Committee views
departures from the inflation target induced by
temporary supply shocks.
Still, the choice between core and overall measJ U LY / A U G U S T 2 0 0 4

157

REVIEW

Meyer

ures of inflation would make a difference in the
conduct of monetary policy, at least if policymakers
were responding to recent changes in inflation in
their decisions about the setting of the funds rate
target. As a result, the optimal response to price
shocks remains an important consideration in the
choice between core and overall inflation rates as
targets. If it is optimal to “look through” the direct
effects of a price shock, and only respond to the
extent that there are indirect effects that later raise
the core inflation rate, this might suggest a preference
for the core measure. On the other hand, the presumption that some portion of a price shock would
likely pass-through to the core may suggest the desirability of some initial response to the direct effect.

Should the Target Be Set as a Point or
a Range, and, if a Range, Should There
Be a Special Focus on the Midpoint?
I prefer either a point target or a range with a
focus in the midpoint as the explicit target. This
would likely provide a better anchor for inflation
expectations and reduce the indecision in the markets when the central bank was at one end of the
range about whether or not the central bank would
look for an opportunity to move back to the middle.
That does leave the question of what the purpose
of a range is and how wide the range should be.
One purpose is to identify a range of variation that
is typical cyclically and would not be as strongly
resisted as when inflation moves outside the range.
That suggests a kind of nonlinear policy response
that could, in turn, be effective in limiting the variation of inflation expectations. The range might also
identify, for example, the upper limit to where the
Committee would be comfortable pursuing an
opportunistic disinflation strategy, and the lower
limit could identify the level below which policy
would be more focused on erring on the side of
ease, because of concerns about the possibility of
deflation or hitting the zero nominal bound.
Many inflation-targeting countries have chosen
a range of 1 percentage point, typically from 1
percent to 3 percent. Governor Bernanke has indicated that he would like to see inflation within a 1percentage-point range, 1 percent to 2 percent.

Should the Inflation Target Be Set
Once and for All or Be Subject to
Adjustment?
The spirit of an inflation target is that it should
be set and remain in place for long periods, so as to
158

J U LY / A U G U S T 2 0 0 4

ensure economic agents that they can make longerrun decisions with confidence about the average
inflation rate over such horizons.
But, while the target should not be changed
often, there should be a willingness to revisit the
target, on occasion, as evidence about the inflation
bias evolves and as research provides new information about the appropriate size of the cushion relative to zero true inflation.

Should the Target Be a Price Level or
Inflation Target?
There has been considerable discussion about
the benefits of a price level rule for an economy
facing the danger of deflation. Similar benefits accrue
to a target for the average inflation rate over some
period, as long as there is a commitment to compensating for periods when inflation is below the target
with periods where inflation is above the target. However, as I have suggested, the case for moving to an
explicit numerical inflation target is generally perceived to be an attempt to preserve continuity in U.S.
monetary policy, not to provide an opportunity for
a significant change in the way in which that policy
is conducted. So I would not anticipate that an option
of a price level target would be seriously considered.

Should the Inflation Target Be Set for
True Price Stability or Price Stability
Plus a Cushion, and, if There Should
Be a Cushion, How Large Should It Be?
The FOMC considered this topic in considerable
detail at the first FOMC meeting I participated in, in
July 1996. Janet Yellen made the case for an inflation
target set high enough to both take into account
measurement error and also allow a cushion that
took into account the potential deterioration in
economic performance if inflation were too low.
She called for an inflation target of 2 percent and
everyone on the Committee lined up to state their
preference. The Chairman tried to get away with his
vague definition of price stability: “Price stability is
that state in which expected changes in the general
price level do not effectively alter business or household decisions.” But Yellen pressed him and asked
if he could put a number on that. Remarkably, the
Chairman agreed, and said he preferred zero inflation, correctly measured. Janet asked if he could
settle for 2 percent incorrectly measured.
By the way, this is the only time during my 51/2
years on the Board and the FOMC that anyone was
able to extract from the Chairman a number related

FEDERAL R ESERVE BANK OF ST. LOUIS

to his forecast, his estimate of productivity growth
or anything else—of course, other than his recommendation each meeting for the federal funds rate
target.
During a go-around on the topic, only a few
Committee members preferred a target of zero, and
the consensus was very strong for a 2 percent inflation target. The Chairman ended up summarizing
the discussion as “an agreement for 2%,” but then
cautioned Committee members not to reveal that
such a discussion even took place.
Interestingly, the Chairman asked toward the
end of the discussion to what measure of inflation
the 2 percent target should apply. Yellen indicated
she did not have a specific measure in mind, but
most of the Committee appeared to be thinking in
terms of the CPI, specifically the core CPI. Greenspan
argued that the PCE was the better measure of consumer price inflation and that the target should be
set in terms of the best measure. He then pointed
out that while the core CPI was 21/2 percent, the core
PCE was already 2 percent, so that the Committee
could apparently declare victory.
Bob McTeer noted, however, that the specific
target depends on the specific measure. If the
Committee preferred 2 percent for the core CPI, the
consistent target for the core PCE would be 11/2
percent, given the recent differentials among the
measures. Interestingly, the Chairman’s apparent
acquiescence to a 2 percent target for the core PCE
would have left him with a higher target for inflation
than preferred by the rest of the Committee.
What if that discussion were opened up today?
Governor Bernanke has indicated his preference
for a target of 1 to 2 percent, presumably for the
PCE measure, in line with the spirit of the July
1996 meeting. But a lot has happened since then,
particularly experience in Japan with deflation and
in the United States with low inflation.
The lessons drawn from these experiences have
reinforced the wisdom of Yellen’s remarks in July
1996, specifically that inflation can be too low as
well as too high, and that monetary policymakers
need to raise inflation to its target when inflation
falls below the target, just as they need to lower
inflation when it rises above the target.
Indeed, the lessons from recent experience
suggest that policy should be asymmetric, in light
of the asymmetric risks associated with deflation
and the zero nominal bound. That is, policymakers
should be more aggressive raising inflation to its
target when it is initially too low than lowering it to
its target when it is initially too high.

Meyer

An interesting question is whether the inflation
target should be set high enough so that policymakers could respond symmetrically to movements
in inflation above and below the target, except perhaps in a small percentage of cases.
That suggests that consideration might be given,
for example, to a 11/2 percent or 2 percent target
for the core PCE.

Should Additional Reporting Requirements Accompany the Introduction of
an Explicit Numerical Inflation Target?
A feature of inflation-targeting regimes, in
additional to an explicit numerical inflation target
and a hierarchical mandate, is greater transparency
about the forecast and a greater focus on explaining
any departures from the target.
First, the FOMC should not issue a separate
“inflation report”—because that would be inconsistent with the spirit of the dual mandate. The only
change relative to the Monetary Policy Report and
semi-annual testimony would be some explicit
commentary on the outcome for inflation relative
to the target, and, when inflation is outside the
monitoring range, why that occurred and how the
Committee viewed the process and timetable for a
return to an inflation rate inside the range.
Second, the FOMC forecast should explicitly
include whatever inflation measure the target is
based upon. Today, the FOMC provides its forecast
of the overall inflation rate for the PCE, while many,
including myself, believe that the Committee makes
its decisions based more on the core measure. If
the target is stated in terms of the core measure, it
should be included in the FOMC forecast.
Third, since the FOMC controls inflation over
the intermediate term, it would be useful if the FOMC
forecasts always went out at least 11/2 to 2 years.
The current practice is that the FOMC forecast in
late January or early February only extends through
the remainder of that year. This should be extended
for another year.
Fourth, it might be useful to increase the frequency of FOMC forecasts from twice per year to
four times. Just as a picture is worth a thousand
words, a forecast can be more revealing than
speeches and testimonies.
Fifth, if there is more attention on the forecast,
the Committee should fine-tune the process by
which they are prepared by the individual Committee members. The forecasts are supposed to be
based on “appropriate” monetary policy, but the
J U LY / A U G U S T 2 0 0 4

159

REVIEW

Meyer

forecasts the staff provides the Committee are often
based on a constant nominal funds rate. The staff
should help the Committee members in the forecast
process by always providing them a forecast based
on a policy rule or an “optimal policy” simulation.

AN EXAMPLE OF AN EXPLICIT
NUMERICAL INFLATION TARGET
The FOMC will conduct policy in an effort to
achieve maximum sustainable employment and
price stability, where the latter is defined as an
inflation rate of 11/2 percent, measured by the core
PCE inflation rate. Given that the economy is subject
to shocks and business cycles, it will be impossible
for the Committee to achieve simultaneously both
objectives at each moment in time. The objective
for employment is to minimize the variance of
employment relative to its maximum sustainable
level. The objective for price stability is to achieve
an average for the rate of inflation as close as possible to the inflation target over the business cycle.
The inflation target is symmetric, as the Committee recognizes that inflation can be too low as
well as too high. Therefore, monetary policy would
be directed to raising inflation if it fell below the target and lowering inflation if it rose above the target.
The Committee has intentionally set the target
so as to provide some cushion to reduce the likelihood that the economy could encounter deflation
or that the federal funds rate could reach the zero
nominal bound.
The Committee, in conducting its policy, finds
it useful to establish a monitoring range around the
inflation target of 1 to 2 percent. Movements within
the band represent cyclical variation in the inflation
rate that is acceptable to the Committee. Movements
outside this range would be of more concern and
would be more vigorously countered by monetary
policy, with appropriate consideration as to whether
the divergences were likely to be temporary or more
permanent, without the intervention of policy.

160

J U LY / A U G U S T 2 0 0 4

Commentary
Lars E.O. Svensson

L

arry Meyer has presented us with a fine paper
on practical problems and obstacles to inflation targeting in the United States—a very
suitable paper for this great conference at the St.
Louis Fed. I find it a very thoughtful paper with
many good points. I agree with most of the points
Larry makes; I hope many people will read the
paper and, in particular, take seriously his proposal
for inflation targeting in the United States.
However, one issue that I believe Larry puts too
much weight on is his pet idea, the distinction
between a dual and hierarchical mandate. I do not
believe this distinction is very useful. We are all
flexible inflation targeters now. More precisely, we
are all talking about an intertemporal loss function
for monetary policy consisting of the expected discounted sum of present and future period losses, that
is, of the form
`

Et ∑ (1 − δ )δ τ Lt + τ ,
τ =0

where the period loss function is typically given by
Lt =

[

]

1
2
2
π t − π *) + λ y ( yt − yt ) .
(
2

Here, Et denotes expectations conditional on information available in period t (typically a quarter), δ
is a discount factor and fulfills 0<δ<1, πt denotes
an inflation measure in period t (typically fourquarter inflation for a specified price index), π*
denotes the inflation target, yt denotes log output,
y–t denotes log potential output, yt – y–t is the output
gap, and λ y>0 is the relative weight on output-gap
stabilization relative to inflation-gap stabilization.
Alternatively, the period loss function can be
expressed in terms of employment,
Lt =

(

)

2
1
2
π t − π *) + λ l lt − lt  ,
(


2

–
where lt and l t denote log employment and log
–
equilibrium employment, respectively, lt – l t is the
employment gap, and λ l>0 is the relative weight
on employment-gap stabilization.1
Let us look at the loss function in terms of the
output gap. There are three parameters there: δ,
π*, and λ y. The δ, the discount factor, is very close
to 1 and not a big issue. The π* is the inflation target,
announced explicitly by an inflation targeter. There
remains only one parameter, the λ y, the relative
weight on output-gap stabilization. The adjective
“flexible” in “flexible inflation targeting” has to do
with the value of λ y.
The issue is really how to describe a loss function
of this type in words rather than a formula. I do not
believe the dual-hierarchical distinction helps in this
respect. But, if we must use it, we can think of this
loss function as saying something about the first
moments, the long-run means, of the target variables,
inflation and output; and the second moments, the
variability of the target variables around those means.
Regarding the first moments, there is a target
for long-run mean inflation, π*. This target is subject
to choice by the central bank or by its principal, the
government or the parliament, depending on the
institutional setup in the country. But the target for
output is not subject to choice. It is a “fact.” It is given
by the economy, by its potential output. Since potential output is an unobserved variable, it requires
estimation. We can say that the output target is subject to estimation, but it is certainly not subject to
choice. Alternatively, we can say that the output-gap
target is given at zero, and also not subject to choice.
Since there is a meaningful choice of the target for
long-run inflation but not of the target for long-run
1

The period loss function could also be expressed in terms of the
– , instead of the employment gap, where u
unemployment gap, ut – u
t
t
– denotes equilibrium unemployment.
denotes unemployment and u
t

Lars E.O. Svensson is a professor of economics at Princeton University, a research fellow of the Centre for Economic Policy Research, and a research
associate of the National Bureau of Economic Research. The author thanks Norbert Janssen for editorial assistance.
Federal Reserve Bank of St. Louis Review, July/August 2004, 86(4), pp. 161-4.
© 2004, The Federal Reserve Bank of St. Louis.

J U LY / A U G U S T 2 0 0 4

161

REVIEW

Svensson

output or the output gap, and therefore an asymmetry between the two targets, we can, if we like, talk
about a hierarchical mandate for long-run inflation.
Regarding the second moments, the parameter
λ y expresses the weight on the loss from output-gap
variability relative to the loss from inflation variability. Whenever λ y is positive, output-gap stability, as
well as inflation stability, is an objective. There is a
symmetry between the two variability objectives,
and we can, if we like, talk about a dual mandate for
inflation-gap variability and output-gap variability.
Since all inflation targeters are flexible inflation
targeters, in the sense that they are concerned about
stability of the real economy in addition to stability
of inflation, we can, if we like, talk about inflation
targeters as having a dual mandate. But, as long as
we know that we are talking about different verbal
descriptions of monetary policy loss functions of the
kind stated above, I do not find the dual/hierarchical
mandate distinction helpful. In particular, it is misleading to say that inflation targeters have a hierarchical mandate but the Fed has not.
Let us compare the mandates for the Fed and
for the Reserve Bank of New Zealand (RBNZ). The
mandate for the Fed, as expressed in the Federal
Reserve Act, is: “[The Fed shall] promote effectively
the goals of maximum employment, stable prices,
and moderate long-term interest rates.” The mandate
for the RBNZ, as expressed in the 2002 Policy Target
Agreement (PTA), is: “[The RBNZ’s policy target] shall
be to keep future CPI inflation outcomes between 1
per cent and 3 per cent on average over the medium
term.” But the PTA also states that, “[i]n pursuing
its price stability objective, the Bank...shall seek to
avoid unnecessary instability in output, interest
rates and the exchange rate.”
First, I do not find the formulation of words in
the mandate for the Fed to be particularly wellchosen. These words need careful interpretation to
be a meaningful mandate for monetary policy. The
expression “maximum employment” does not make
sense from the point of view of modern monetary
macroeconomics. Taken literally, it leads directly
down the inflation-bias lane of Kydland, Prescott,
Barro, and Gordon. Instead, the Fed prefers to deviate
from the words of the Federal Reserve Act by inserting the word “sustainable” before “employment.”
This insertion allows the interpretation of “maximum
employment” as “equilibrium,” “natural,” or “potential” employment, which is required to make sense
of the mandate. This way, the mandate can be said
to describe a loss function like the ones above,
162

J U LY / A U G U S T 2 0 0 4

although without saying anything about the parameters, except that λ l is positive.
Second, I do not see that the New Zealand PTA
says anything substantially different about the form
of the loss function than the (reinterpreted) U.S.
Federal Reserve Act. “Avoiding unnecessary instability in output” can certainly be interpreted as “avoiding unnecessary instability in the output gap” and
be seen as a verbal description of the loss function
above. Thus, the PTA says that the λ y is positive, in
addition to specifying the parameter π* (the latter
we can interpret as the midpoint of the target range,
that is, 2 percent).2
In conclusion, I do not see that these two mandates are different in a way that makes it meaningful
to say that one is dual and the other is hierarchical.
To continue with Larry’s paper, his goals for a
change in the Fed framework are to improve transparency and accountability and to enhance the
effectiveness of monetary policy, without fundamentally altering the basic approach to the conduct
of monetary policy under the Greenspan Federal
Open Market Committee (FOMC). He goes on to
discuss the vision of the Greenspan FOMC and summarizes it in three principles. I believe one could
add a fourth principle to these, namely, “Avoid commitment, transparency, and accountability.” It seems
to me that the vision of the Greenspan FOMC includes
maintaining maximum discretion, including maximum discretion about interpreting and reinterpreting
the mandate. This is “flexibility,” but in a very different—and rather undesirable—sense from the flexibility in flexible targeting.
I believe the world has voted many times on
the Greenspan Fed versus inflation targeting. As far
as I know, no country has copied the vision of the
Greenspan Fed, but many countries have copied
the inflation-targeting framework of the RBNZ, the
Bank of England, and Sweden’s Riksbank. Furthermore, Michael Woodford’s (2004) paper for this
conference explains clearly and convincingly how
commitment and transparency are essential in making more effective the management of expectations,
which is at the core of modern monetary policy,
and, in particular, how this provides a strong argument for inflation targeting. In my own work, for
instance, in Svensson (1999), I have emphasized
2

In my review of NZ monetary policy, Svensson (2001), a major issue
was whether the RBNZ had conducted monetary policy with the
appropriate degree of flexibility, that is, whether it had successfully
avoided unnecessary variability of output, interest rates, and the
exchange rate. My conclusion was that it had.

FEDERAL R ESERVE BANK OF ST. LOUIS

that inflation targeting is fundamentally a commitment to sensible monetary policy objectives and to
transparency about those objectives. This puts inflation targeting in stark contrast to this fourth principle
of the vision of the Greenspan FOMC.
Some have noted that the Fed has become more
transparent in recent years. This is true, but most of
it seems to consist of somewhat reluctant concessions to outside pressure and the example set by the
inflation targeters rather than enthusiastic reforms
from within. I suggest that this fourth principle be
abandoned.
Larry goes on to discuss the politics of inflation
targeting, why to bother about a change in the framework at all, why to do it now, and what the obstacles
are. He also discusses the case against inflation
targeting.
The case against has most explicitly been laid
out by Kohn (2003), with worries about the possible
loss of flexibility: “[T]he success of U.S. monetary
policy has in large part derived from its ability to
adapt to changing conditions—a flexibility that likely
has benefited from the absence of an inflation
target.”3 As far as I can see, the flexibility referred
to here is the flexibility to reinterpret the mandate
and change the monetary policy objectives. I believe
such flexibility is flexibility of the wrong kind.
Instead, I believe that all the flexibility we need is
the flexibility summarized in the λ above, the weight
on stabilization other than that of the inflation gap.
I do not see what prevents the Fed from announcing
an inflation target and becoming an explicit inflation
targeter and, in particular, introducing its own variant of inflation targeting, a high-λ, high-flexibility
one. Indeed, nothing prevents the Fed from being an
explicit super-flexible inflation targeter, if that is what
it wants to be. It may even call itself an “inflationand-output-gap targeter,” or an “inflation-and-realeconomy stabilizer,” or whatever. The Fed might
indeed want to set an example for the world, by
being the most flexible inflation targeter in history.
What I do not get is what the social benefit is from
the Fed being fuzzy about its objectives.4
In sum, there is simply no case against inflation
targeting for the United States. There is no downside
to inflation targeting for advanced countries. This
is furthermore demonstrated by the fact that no
country has, to my knowledge, had any regrets about
3

See McCallum (2003) for a strong rebuttal of Kohn’s arguments.

4

Debelle (2003) provides an interesting description and discussion of
the flexible inflation targeting of the Reserve Bank of Australia.

Svensson

adopting inflation targeting. In contrast, the view
in these countries seems to be that it is the best
monetary policy regime they have ever had.
Larry goes on to discuss a number of practical
considerations, with many good and concrete suggestions. He suggests not calling the monetary policy
report “Inflation Report,” but calling it “Monetary
Policy Report” instead. I have no quarrel with this.
Indeed, many inflation-targeting central banks call
their report something other than “Inflation Report”;
the RBNZ calls it “Monetary Policy Statement.” The
important thing is not the name but that it provides
the appropriate information.
In particular, I believe that the report should
ideally include forecasts of inflation, output, potential output, and the interest rate, with the appropriate
fan charts to indicate the uncertainty in these forecasts. The forecasts should be the best unconditional
forecasts; that is, they should be for the optimal
interest-rate path, the most likely future interest-rate
path. Only then do the forecasts provide the best
guide for private-sector expectations, and only then
does it make sense to compare the forecasts with
ex post outcomes. Furthermore, I believe the forecasts should extend to three years rather than the
standard two (the RBNZ and Bank of Norway already
have three-year forecasts). It is awkward when the
forecasts end at two years, since there is often quite
a bit of discussion of the two-year horizon. In most
cases, there would not be much specific information
about the third year, which implies that in most cases
the forecasts for the third year would be flat on target;
but it will be reassuring to the general public that
there is nothing dramatic lurking beyond the twoyear horizon that the central bank is aware of but
does not mention. Three-year forecasts would also
be more in line with the increasingly frequent reference to “the mid term.” In the interest of increased
transparency, I would humbly suggest that this term
be replaced by “1.5–3 years” or similar, whenever
possible.
A recent innovation of the Bank of Norway—
an enthusiastic newcomer to the inflation-targeting
camp that has moved straight into the group of bestpractice inflation targeters (see Svensson et al.,
2002)—is to plot the inflation forecast and the
output-gap forecast in the same graph (see chart 1
in Norges Bank, 2003). This clearly serves to emphasize that the Bank is concerned with the stability of
the real economy as well as with inflation, emphasizing the flexibility in its inflation targeting.
Finally, Larry provides a concrete example of
J U LY / A U G U S T 2 0 0 4

163

Svensson

an inflation-targeting proposal for the Fed, with an
inflation target of 1.5 percent for the core PCE price
index, in line with the proposal of Goodfriend (2003).
I hope that a proposal similar to these is adopted
soon.

REFERENCES
Debelle, Guy. “The Australian Approach to Inflation
Targeting.” Working paper, 2003.
Goodfriend, Marvin. “Inflation Targeting in the United
States?” in Ben S. Bernanke and Michael Woodford, eds.,
The Inflation Targeting Debate. Chicago: University of
Chicago Press (forthcoming);
www.nber.org/books/inflation-targeting.
Kohn, Donald L. Comment on M. Goodfriend, “Inflation
Targeting in the United States?” in Ben S. Bernanke and
Michael Woodford, eds., The Inflation Targeting Debate.
Chicago: University of Chicago Press (forthcoming);
www.nber.org/books/inflation-targeting.
McCallum, Bennett T. “Inflation Targeting for the United
States.” Working paper, 2003;
wpweb2k.gsia.cmu.edu/faculty/mccallum/ITforUS5.pdf.
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.
Norges Bank. Inflation Report 3/2003. 2003;
www.norges-bank.no.
Svensson, Lars E.O. “Inflation Targeting as a Monetary
Policy Rule.” Journal of Monetary Economics, June 1999,
43(3), pp. 607-54.
Svensson, Lars E.O. “Independent Review of the Operation
of Monetary Policy in New Zealand: Report to the Minister
of Finance.” 2001; www.princeton.edu/~svensson.
Svensson, Lars E.O.; Houg, Kjetil; Haakon, Solheim and
Steigum, Erling. “An Independent Review of Monetary
Policy and Institutions in Norway.” Norges Bank Watch
2002, Centre for Monetary Economics, Norwegian
School of Management BI, 2002;
www.princeton.edu/~svensson.
Woodford, Michael. “Inflation Targeting: A Theoretical
Evaluation.” Federal Reserve Bank of St. Louis Review,
July/August 2004, 86(4), pp. 15-41.

164

J U LY / A U G U S T 2 0 0 4

REVIEW

Panel Discussion

Inflation Targeting
Ben S. Bernanke

S

hould the Federal Reserve announce a quantitative inflation objective? Those opposed to
the idea have noted, correctly, that the Fed
has built strong credibility as an inflation-fighter
without taking that step and that this credibility has
allowed the Fed to be relatively flexible in responding to short-run disturbances to output and employment without destabilizing inflation expectations.
So, the opponents argue, why reduce this flexibility
unnecessarily by announcing an explicit target for
inflation?
It would be foolish to deny that the Fed has
been quite successful on the whole over the past
two decades. Whether the U.S. central bank would
have been even more successful, had it announced
an explicit objective for inflation at some point, is
impossible to say. We just don’t know. We can’t rerun history; and although empirical cross-country
comparisons can be useful, they are far from being
controlled experiments.
However, the relevant question at this point is not
the unknowable outcome of the historical counterfactual but whether, given the initial conditions we
face today, the adoption of an explicit inflation
objective might not improve U.S. monetary policy
in the future. The Fed’s environment today is different from that of the 1980s and 1990s in at least one
important respect: Price stability is no longer just
over the horizon, but has been achieved—core inflation rates are currently not much above 1 percent.
Thus, in contrast to the experience of the past 35
years or so, in which there could be little doubt about
the Fed’s desired direction for inflation, today the
risks to inflation are more nearly symmetrical; that
is, inflation can be too low as well as too high.

A case can be made, I believe, that when the
economy is operating in the region of price stability,
public expectations and beliefs about the central
bank’s plans and objectives, always important,
become even more so. First, because the public
can no longer safely assume that the central bank
prefers lower to higher inflation, expectations
about future policy actions and future inflation
may become highly sensitive to what the public
perceives to be the Fed’s “just right” level of inflation.
Uncertainty about this “just right” level of inflation
thus may translate, in turn, into broader economic
and financial uncertainty. Second, at very low inflation rates, the zero lower bound on the policy interest rate is more likely to become relevant, which
increases the potential importance of effective expectations management by monetary policymakers.
For example, when interest rates are very low, the
best way to ease policy may be to explain to the
public that the current low interest rate will be maintained for a longer period, rather than simply lowering the current rate. It seems to me that the enhanced
importance of public beliefs and expectations about
monetary policy in the region of price stability
argues for greater attention by the central bank to
its methods of communication when inflation is low.
On the premise that effective communication
is even more crucial near price stability, I will focus
today on how an incremental move toward inflation
targeting, in the form of the announcement of a longrun inflation objective, might help the Fed communicate better and perhaps improve policy decisions
as well, without the costs feared by those concerned
about a potential loss of flexibility. As usual, my
views are not to be attributed to my colleagues on
the Board of Governors of the Federal Reserve
System or the Federal Open Market Committee.
As a preliminary, I need to introduce the idea
of the optimal long-run inflation rate, or OLIR for
short. (Suggestions for a catchier name are welcome.)

Ben S. Bernanke is a member of the Board of Governors of the Federal Reserve System.
Federal Reserve Bank of St. Louis Review, July/August 2004, 86(4), pp. 165-68.
© 2004, The Federal Reserve Bank of St. Louis.

J U LY / A U G U S T 2 0 0 4

165

Panel Discussion

The OLIR is the long-run (or steady-state) inflation
rate that achieves the best average economic performance over time with respect to both the inflation
and output objectives.
Note that the OLIR is the relevant concept for
dual-mandate central banks, like the Federal Reserve.
Thus, it is not necessarily equivalent to literal price
stability or zero inflation adjusted for the usual
measurement error bias. Rather, under a dual mandate, a strong case can be made that, below a certain
inflation rate, the benefits of reduced microeconomic
distortions gained from price stability are outweighed
by the costs of too-frequent encounters of the funds
rate with the zero-lower-bound on nominal interest
rates. (This argument underlies the common view
that there should be a “buffer zone” against deflation.) Hence, in general, the OLIR will be greater
than zero inflation, correctly measured. Note also
that the OLIR is an average long-run rate; variation
of actual inflation around the OLIR over the business
cycle would be expected and acceptable (Meyer,
2004).
What is the OLIR for the U.S. economy? A fairly
extensive recent literature has attempted to quantify
the OLIR. (See, e.g., Coenen, Orphanides, and
Wieland, 2003, and references therein.) Because
direct measures of the benefits of low inflation are
not available, papers in this literature, in practice,
estimate the OLIR to be the lowest inflation rate for
which the risk of the funds rate hitting the lower
bound appears to be “acceptably small.” Interestingly,
the results using this approach seem fairly consistent
across models and specifications, with several papers
(including work using FRBUS, the Federal Reserve’s
large econometric model of the U.S. economy; see
Reifschneider and Williams, 2000) having concluded
that the risk of hitting the zero bound seems to
decline sharply once the long-run average inflation
rate rises to about 2 percent. In addition, other
studies of the costs of very low inflation (such as
the supposed effects of downward nominal wage
rigidity on the allocation of labor) have found that
these costs are also largely eliminated at inflation
rates of about 2 percent (Akerlof, Dickens, and Perry,
1996; see also Altig, 2003).
Fortuitously, then, it may be the case that something in the vicinity of 2 percent is the optimal longrun average inflation rate for a variety of assumptions
about the costs of inflation, the structure of the
economy, the distribution of shocks, etc. However,
before we embrace that number, many details
remain to be filled in. For example, in practice, much
166

J U LY / A U G U S T 2 0 0 4

REVIEW
might depend on the specification of the inflation
index, on assumptions made about the steady-state
value of the real interest rate, and on other factors.
Also important would be getting a better sense of
the range of uncertainty around this number. More
research on this issue would be highly worthwhile.
As the economy seems currently to be moving
toward a sustainable expansion path, with a stabilizing rate of inflation, having an estimate of the OLIR
likewise seems crucial to making good policy in the
next few years. The issue is one that, in my view, the
FOMC and the staff should be looking at carefully.
Suppose, as I believe would be feasible, that the
FOMC were able to agree on a value or central tendency for the OLIR, based on the results of staff
research and discussion among Committee members. Of course, the value of the OLIR would only
be a rough approximation to the “truth,” but one
cannot avoid making such approximations in policymaking, whether implicitly or explicitly. Should the
FOMC then take the next step and announce this
number to the public? Some have argued that such
an announcement would be unnecessary because
the Fed’s implicit inflation objective is already well
understood by the market. I am skeptical. Publicly
expressed preferences by FOMC members for longrun inflation have ranged considerably, from less
than 1 percent to 2.5 percent or more. Long-run
inflation expectations implicit in the pricing of
inflation-indexed securities vary significantly over
time, and the apparently high sensitivity of long-term
nominal interest rates to Fed actions suggests some
uncertainty about the Fed’s long-run inflation target
(Gurkaynak, Sack, and Swanson, 2003). Gavin (2004)
points out that the range of private-sector forecasts
for inflation is typically higher for the United States
than for inflation-targeting countries.
If announcing the OLIR does not constrain shortrun policy unduly, I really cannot see any argument
against it. To reassure those worried about possible
loss of short-run flexibility, my proposal is that the
FOMC announce its value for the OLIR to the public
with the following provisos (not necessarily in these
exact words):
(i) The FOMC believes that the stated inflation
rate is the one that best promotes its output,
employment, and price stability goals in the
long run. Hence, in the long run, the FOMC
will try to guide the inflation rate toward the
stated value and maintain it near that value
on average over the business cycle.

FEDERAL R ESERVE BANK OF ST. LOUIS

(ii) However, the FOMC regards this inflation rate
as a long-run objective only and sets no fixed
time frame for reaching it. In particular, in
deciding how quickly to move toward the
long-run inflation objective, the FOMC will
always take into account the implications for
near-term economic and financial stability.
As you can see, stating the OLIR with these
provisos places no unwanted constraints on shortrun monetary policy, leaving the Committee free to
deal with current financial and cyclical conditions as
the Committee sees fit. In this respect, the proposal
is very similar to one recently advanced by Governor
Gramlich (2003).
To be clear, because neither the horizon at
which the inflation objective is to be attained nor
the expected path of inflation and output is specified
under this proposal, what I am suggesting is not
equivalent to inflation targeting as commonly understood. Instead, what is being proposed is an incremental step that I believe would provide important
benefits in itself and which would leave the door
open for further steps later if that seemed appropriate. In the language of Faust and Henderson (2004)
at this conference, my objective is to get the mean
of inflation right while leaving the determination of
the variance open for future discussion and debate.
Without any fixed time frames for reaching the
optimal long-run inflation rate, would an announced
value for the OLIR carry any credibility? I think it
would, for the important reason that the OLIR is not
an arbitrarily selected value. In particular, because
this inflation rate would have been judged by the
Committee to be the one under which the economy
operates best in the long run, the FOMC would have
an incentive to try to reach it eventually, even if it
were not an announced long-run objective of policy.
Thus, despite the lack of a time frame, the OLIR
should have long-run credibility; that is, it should
be the best (lowest-forecast-error) answer to the
question, What do you expect the average inflation
rate in the United States to be over the ten-year
period that begins (say) three years from now?
Additional reasons that the announcement
would carry weight are the accumulated credibility
of the Fed and the fact that we are presumably starting from a point near the optimal inflation rate, so
that a period of costly disinflation will not be needed
to reach the OLIR. In other words, this relatively
unconstrained approach might not work for other
central banks, and it might not have worked for the

Panel Discussion

Fed at other times (e.g., when we were at early stages
of the disinflation process); but given the current
configuration of circumstances, it should work now.
I have argued that announcing the OLIR would
not have significant costs. What are the benefits?
In my view, the announcement of the OLIR should
serve as a useful clarification of the long-run objective of the Fed and would thereby provide a long-run
“anchor” to monetary policy. Among other benefits,
the announcement of the OLIR should help participants in financial markets price long-term bonds
and other financial assets more efficiently; help to
lower inflation risk in financial markets and in other
forms of contracting; and tend to stabilize long-term
inflation expectations more broadly, which in turn
would make short-run stabilization policy more
effective (Orphanides and Williams, 2003). Although
the announcement of the OLIR would not constrain
short-run policymaking in undesirable ways, it would
nevertheless also help the market make inferences
about the likely timing and extent of tightening and
easing cycles, since, all else equal, the FOMC would
want the inflation rate to move “asymptotically”
toward the long-run desired level. For example, if
the current inflation rate were known to be below
the OLIR, that fact would convey some information
about how long it will likely be before the Fed begins
its tightening cycle.
Because some of the principal benefits of
announcing the OLIR would arise from the reduction
of uncertainty in financial markets and in the economy more broadly, I prefer the announcement of a
single number for the OLIR, or at least a number
with a surrounding tolerance range that is as narrow
as the Committee can live with. I acknowledge that
the OLIR cannot be determined precisely. Nevertheless, to the extent that the FOMC is fairly indifferent
over a modest range of long-run inflation rates, there
would be a positive benefit to choosing a single
number within that range and trying to coordinate
public expectations on that number.
Agreeing on and announcing a value for the
OLIR might improve policymaking more directly,
at least on the margin. In particular, the stated inflation objective would help guide policy during periods, like now, in which the economy is (we hope)
returning to a sustainable growth path; at all times,
it would also serve as a reminder to policymakers
to keep one eye on the long run at the same time
that they are reacting to current developments in
the economy. But, to reiterate, it seems likely that the
biggest gains would be in the area of communicaJ U LY / A U G U S T 2 0 0 4

167

Panel Discussion

tions. Sharing the OLIR with the public would
address the most important information asymmetry
in the system: namely, the public’s imperfect knowledge of the FOMC’s objectives. I believe this step
would help to reduce the reliance of the Fed on complex and easily misinterpreted qualitative language
in its communications with the public.
I conclude with a word on the politics of this
proposal. One concern frequently expressed about
announcing an inflation objective is that the Congress
would interpret the introduction of an inflation target
as a repudiation of the dual mandate. This would
be a misinterpretation, but I understand why some
legislators might draw the wrong conclusion.
However, it seems to me that the recent attention
to the risk of deflation changes the political calculus.
There now exists a broad awareness that an inflation
rate that is too low, by raising the probability of
deflation and a binding zero bound on the nominal
interest rate, poses a threat to output and employment stability. Therefore the connection between
the announced OLIR and the real side of the economy will be much more apparent to non-economists.
Indeed, the entire rationale for the OLIR can be
expressed in terms of jobs and growth. The FOMC
might say to Congress: “We don’t want long-run
inflation to be too high, because low inflation promotes growth and productivity. On the other hand,
inflation shouldn’t be too low, because we want to
have all the room we need to respond to the dangers
that deflation poses for output and employment.
We pose the objective in terms of inflation only
because that is what the Fed can control in the long
run.” It does not seem to me to be such a difficult
case to make in terms of the existing dual mandate.
In addition we would have the explicit proviso that
important short-run economic and financial goals
will not be sacrificed to reach the long-term inflation
objective more quickly. Although it would be important to vet these ideas thoroughly with the relevant
Congressional committees before proceeding, I am
hopeful that a change of the type I am proposing
would be acceptable to Congress as being within
the spirit of existing legislation.

REFERENCES
Akerlof, George A.; Dickens, William T. and Perry, George L.
“The Macroeconomics of Low Inflation.” Brookings
Papers on Economic Activity, 1996, (1), pp. 1-76.

168

J U LY / A U G U S T 2 0 0 4

REVIEW
Altig, David. “What Is the Right Inflation Rate?” Federal
Reserve Bank of Cleveland Economic Commentary,
September 15, 2003.
Coenen, Günter; Orphanides, Athanasios and Wieland,
Volker. “Price Stability and Monetary Policy Effectiveness
when Nominal Interest Rates Are Bounded at Zero.” ECB
Working Paper No. 231, European Central Bank, May 2003.
Faust, Jon and Henderson, Dale W. “Is Inflation Targeting
Best-Practice Monetary Policy?” Federal Reserve Bank of
St. Louis Review, July/August 2004, 86(4), pp. 117-43.
Gavin, William. “Inflation Targeting: Why It Works and
How To Make It Work Better?” Business Economics, April
2004, 39(2), pp. 30-37.
Gramlich, Edward. “Maintaining Price Stability.” Remarks
before the Economic Club of Toronto, Toronto, Canada,
October 1, 2003.
Gurkaynak, Refet S.; Sack, Brian and Swanson, Eric. “The
Excess Sensitivity of Long-Term Interest Rates: Evidence
and Implications for Macroeconomic Models.” Finance
and Discussion Paper 2003-50, Board of Governors of
the Federal Reserve System, November 2003.
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.
Orphanides, Athanasios and Williams, John C. “Imperfect
Knowledge, Inflation Expectations, and Monetary Policy.”
Finance and Discussion Paper 2002-27, Board of
Governors of the Federal Reserve System, June 2003.
Reifschneider, David and Williams, John C. “Three Lessons
for Monetary Policy in a Low-Inflation Era.” Journal of
Money, Credit, and Banking, November 2000, 32(4, Part 2),
pp. 936-72.

FEDERAL R ESERVE BANK OF ST. LOUIS

Panel Discussion

Inflation Targeting:
A View from the ECB

Thus, in section 2, I would like to put the current
debates on monetary policy into a historical perspective. In section 3, I will discuss what I see as the
critical aspects of the inflation-targeting approach.
Section 4 outlines the ECB’s monetary policy
approach and the ways in which it resembles and
differs from the inflation-targeting framework.

Otmar Issing

1. INTRODUCTION

W

hat is the ultimate objective of monetary
policy? What is the appropriate framework for conducting monetary policy?
Central bankers and academics have been asking
these critical questions for decades. This conference,
in which I was honored to take part, was a milestone
in this long-standing debate.
It is a fact that never before in the history of
fiat money has there been so much consensus on
the benefits of a low-inflation environment, and
many central banks have achieved results consistent
with this conviction. This is a tremendous achievement and one that could easily lead us to think that
at last this long-standing debate has been settled
once and for all.
However, I do sometimes wonder whether we
are not too complacent in believing that the regime
of low inflation will be with us “from here to eternity.”
There is always a risk that even great achievements,
after a while, are taken as given and that their value
is only rediscovered when they are in danger of
becoming lost. In addition, recent history should
tell us that the structure of the economy changes
over time in a way that is difficult to anticipate and
perceive in real time. This continuous mutation
makes the task of monetary policy and its implementation even more challenging. It is the intrinsic
nature of the economy that makes the debate on
the aims of monetary policy and its appropriate
framework so difficult to settle, and I believe that
this debate will continue for some time to come.
In the course of the 1990s the inflation-targeting
framework for the conduct of monetary policy
has become popular among central banks and
academics. In this paper I will highlight some of what
I think are the distinguishing features and possible
pitfalls of this approach. I will then draw comparisons with the European Central Bank’s (ECB) monetary policy strategy, also in the light of the ECB’s
clarification of its strategy in May 2003.

2. A HISTORICAL PERSPECTIVE
Following the philosophy of “rules above
authorities”—to paraphrase slightly the title of
Henry Simons’ famous article (1936)—one strand
of research wanted the behavior of central bankers
to be strictly constrained by a rule for conducting
monetary policy. The most prominent advocate of
this was Milton Friedman and his famous k-percent
rule. The key argument in favor of the adoption of
a simple strict rule was the acknowledgment of the
economists’ and central bankers’ ignorance of the
exact functioning of the economy and the long and
variable time lags of monetary policy. It maintains
that the actors of monetary policy know too little
of the actual functioning of the economy to be able
to perform activist policy and their discretionary
actions would only exacerbate economic fluctuations
instead of smoothing them. Strict rules prevent such
problems, eliminating judgmental elements in monetary policy action and avoiding activist policy.
However, during the 1960s, few central
bankers were in favor of rules, mostly because the
performance of discretionary monetary policy at
that time had been quite satisfactory, at least in the
United States, and policymakers were increasingly
confident of their ability to properly steer the performance of the economy. The 1970s marked the
end of that overconfidence. In the period between
the first oil shock and the early 1980s, the world’s
major economies experienced two recessions while
inflation rose to double-digit levels. Although these
events were not fully under the control of the monetary authorities, it is clear that the discretionary
approach to monetary policy did make a negative
contribution by not properly anchoring inflation
expectations and instead allowing them to drift.1
1

Among others, Orphanides and Williams (2003) show how the interaction of policy errors and endogenous expectation formation contributed to stagflation in the 1970s.

Otmar Issing is a member of the Executive Board of the European Central Bank. The author thanks Filippo Altissimo for his valuable contribution.
Federal Reserve Bank of St. Louis Review, July/August 2004, 86(4), pp. 169-79.
© 2004, The Federal Reserve Bank of St. Louis.

J U LY / A U G U S T 2 0 0 4

169

REVIEW

Panel Discussion

One of the lessons that economists learned from
their experience of the 1970s was that economic
agents’ expectations cannot be taken as given by
policymakers when choosing their policy action.
The underlying idea is simple but path-breaking
and goes back to at least Marschak (1947), although
the strongest case was made by Lucas (1976). In
forming their expectations and taking their actions,
economic agents will always try to anticipate future
policy moves. This makes expectations of future
policy relevant for today’s consumption and investment decisions and creates the room for strategic
interaction among economic agents, a cornerstone
of which is the credibility of the policymaker to
commit to a given set of actions. In the context of
monetary policy, Kydland and Prescott (1977) and
Barro and Gordon (1983) proposed models where
the desire of the central bank to attain an unemployment rate below the natural rate generates surprise
inflation in the economy: This is the “time consistency” problem. Economic agents properly understanding the incentives of the monetary authority
and its actions would thus anticipate future inflation.
In equilibrium, this would end up generating the
well-known “inflation bias.” A superior outcome
could be achieved if monetary policy authorities
took into account the effect their behavior could
have on economic agents’ action and properly
commit not to inflate. The advantage of commitment
relative to discretion crucially hinges on the credibility of the monetary authority actually sticking to
its promises.
From those original contributions a large strand
of literature tried to devise incentive-compatible
institutional schemes capable of enforcing a ruletype behavior and thus dealing with the time inconsistency problem. General consensus has emerged
that a necessary prerequisite for solving the time
inconsistency problem is the establishment of an
independent central bank to which the management
of monetary policy is then delegated. The institutional arrangement mostly adopted to enforce the
commitment accepts that monetary policy should
treat the natural rate of unemployment as a given,
and not try to push unemployment below its natural
rate.2
These results square with another finding of the
1970s, namely, the absence of any long-run tradeoff between unemployment and inflation. This point
was stressed by Friedman in his 1977 Nobel lecture,
2

See Walsh (1998) for a survey.

170

J U LY / A U G U S T 2 0 0 4

among others. Friedman’s argument was that, while
it is possible to stimulate the economy in the short
run by some form of monetary illusion, workers
would see through the illusion in the longer run,
demanding higher wages and so bringing employment back to its natural level. Every effort to permanently push employment above its natural level
is therefore self-defeating.
These arguments reinforced the original criticism
of discretionary monetary policy and were the
final nails in the coffin of the theory of an activist
monetary policy (and the idea of a monetary policy
seeking to push economic activity above its natural
level). The focus of monetary policy action had to
be price stability.
The awareness of the limitations of monetary
policy was also coupled with a better understanding
of the possible costs of inflation and the recognition
that a low-inflation environment is a necessary
precondition for long-run growth and an efficient
allocation of resources.3
Taken together, the awareness of the cost of
inflation, of the absence of a long-run trade-off
between inflation and real activity, and of the relevance of the credibility problem of the monetary
authority are some of the motivations underlying
the widespread adoption of a culture of price stability among the central banks of the industrialized
countries during the 1980s and 1990s. I have no
doubt that this new culture has made an important
contribution to the disinflation process that we have
observed in many countries over the past two
decades.4
The inflation experience of the 1970s and
developments in the theory of monetary policy
analysis over the past 20 years have made clear the
importance of the monetary authority making a
firm commitment. However, contrary to the debate
of the 1960s, it is a commitment on an objective
rather than on a simple rule. Once an agreement
on the objective had been reached, another critical
question remained: Which is the best strategy for
achieving this final objective? Over the years central
bankers and academics around the world have
3

For references on the theoretical and empirical literature on the cost
of inflation, see Issing (2001) and Rodriguez-Palenzuela, CambaMendez, and Garcia (2003).

4

Citing the words of another member of the discussion panel of this
conference: “A number of factors have contributed to the reestablishment of price stability, but surely an essential ingredient has been
the attention that the Federal Reserve has paid to long-run trends in
inflation and inflation expectations since 1979” (Kohn, forthcoming).

FEDERAL R ESERVE BANK OF ST. LOUIS

proposed a variety of strategies. Different central
banks have adopted strategies that place different
emphasis on the various pieces of information, or
elements of their decisionmaking process, or different aspects of their communication policies.
Inflation targeting is one of those strategies.
Following the pioneering approach of the Reserve
Bank of New Zealand in the early 1990s, a large
number of central banks have formally adopted an
“inflation targeting framework”; and today we can
count around 20 central banks that refer to this
approach. At the same time the inflation-targeting
framework has triggered a large amount of interesting and stimulating theoretical work, as indeed this
conference testifies.5 Looking back at the experience
of those central banks, there is no doubt that it has
been a success. This is particularly evident in the
case of countries starting from high levels of inflation.
These countries needed to implement a disinflationary process, where inflation targeting served to guide
inflation expectations and provide an explicit framework and direction to monetary policy. The approach
has also turned out to be successful in countries
with lower inflation, as, for example, the positive
experience of the United Kingdom, Sweden, and
Canada shows. In the few cases—limited to some
emerging economies—where the experience has
been somewhat less successful, it is quite evident
that problems originated in other areas—notably,
often stemming from misguided fiscal policies.6
At the same time, while not adopting an inflationtargeting approach, some major central banks have
also achieved and maintained price stability, proving
that visible success in the management of monetary
policy is not confined to inflation-targeting central
banks.
In the rest of this paper I will try to substantiate
this claim.

3. INFLATION TARGETING
There is a vast amount of literature on inflation
targeting, and the first challenge to some readers’
eyes is to decide upon a proper definition. Different
authors have proposed different, and in some cases
conflicting, definitions.7
5

For references on inflation targeting, see, among the others, Bernanke
et al. (1999) and Svensson (1997, 1999, 2000, 2003).

6

See Sims (forthcoming) for an example that a sound fiscal policy is a
prerequisite for the performance of an inflation-targeting framework.

7

See, for example, the two definitions proposed by Amato and Gerlach
(2002) and Svensson (2002) in the same volume of the European
Economic Review.

Panel Discussion

The first and broadest definition of inflation
targeting is simply a monetary policy framework
that accords overriding importance to the maintenance of price stability, typically defined as a low
and stable rate of consumer price inflation.8 As
pointed out in the previous section, given the broad
consensus that price stability is the appropriate
goal of monetary policy, the strategies pursued by
most central banks, including the ECB, would fall
under this loose definition. However, this definition
suffers from two interrelated weaknesses. First, from
the policymaking perspective, it offers no practical
guidance for the conduct of monetary policy beyond
identifying the primary objective. As such, its practical relevance is rather limited. Second, from a
scientific perspective, the definition imposes few
empirically testable restrictions on the implementation of monetary policy. As such, it does not allow
inflation-targeting strategies to be distinguished
from other stability-oriented strategies and their
relative merits to be evaluated. Central banks that
have pursued strategies other than inflation targeting
cannot be meaningfully distinguished on the basis
of this definition. For example, Deutsche Bundesbank
has been classified as an inflation-targeting central
bank by some, despite its long adherence to an
intermediate monetary targeting strategy. To put it
more provocatively, by this definition all “successful”
central banks are inflation targeters, while all
“unsuccessful” central banks are not.
Given the problems associated with this broad
definition, in the remainder of this paper I will focus
on alternative, more restrictive definitions of inflation targeting. Consistent with the existing academic
literature on monetary policy, such narrower definitions are typically expressed in terms of a monetary policy framework based on the adoption of a
monetary policy rule in which forecasts of future
inflation play a central role, either in the form of
the so-called instrument rules or of target rules.
An instrument rule expresses the monetary
policy instrument—usually a short-term nominal
interest rate—as a simple and usually linear function
of deviation of a few key macroeconomic variables,
generally inflation and the output gap, from their
target levels. Usually the literature distinguishes
between an outcome-based rule (if the instrument
is a function of currently observable variables, as
8

Bernanke and Mishkin (1997, p. 97) write: “[Inflation targeting] is
characterized, as the name suggests, by the announcement of official
target ranges for the inflation rate…and by explicit acknowledgement
that low and stable inflation is the overriding goal of monetary policy.”

J U LY / A U G U S T 2 0 0 4

171

REVIEW

Panel Discussion

in Taylor, 1993) and a forecast-based rule (if the
instrument is an explicit function of the current
forecast for key variables in the future).
Under a target rule, the appropriate setting for
the monetary policy instrument is defined implicitly
as the solution to an optimization problem facing
the central bank. This optimization problem is
defined by two elements: first, an explicit loss function describing the costs associated with deviations
of specific goal variable(s) from their target levels;
and, second, a structural model of the economy.
Minimization of the loss function subject to the constraints imposed by the economy’s structure (as
captured by the model) implicitly defines a modelspecific optimal interest rate reaction function, which
determines the interest rate as a function of all
relevant state variables. In this context, an inflationtargeting framework is characterized by the adoption
of a loss function that focuses on the deviations of
forecast inflation from a target level.9
There is a natural complementarity between
instrument and target rule characterizations of
inflation targeting. A target rule implicitly defines
an instrument rule—albeit typically one that is complex and therefore difficult to use in presenting
policy decisions to the public. Similarly, it is usually
possible to derive a loss function and an economic
model that would broadly support a specific instrument rule as the solution to an optimization problem
facing the central bank.
Here, I do not want to enter the vast debate on
the different definitions of and the choice between
instrument and target rules.10 Nor will I address many
of the problematic issues identified by the literature
and associated with the adoption of those rules, such
as the indeterminacy of equilibria, the issue of
commitment to the rules, and the important aspect
concerning the measurement of key variables, for
example, the output gap.11 Instead what I wish to
discuss here are two more practical pitfalls associated
9

This definition corresponds to what is now usually labeled as “strict
inflation targeting.” In “flexible inflation targeting,” on the other hand,
the loss function of the monetary authority focuses on both deviations
of inflation and output from their targets.

10

McCallum (2001b) adds the following disclaimer: “In fact, I believe that
my terminology is more consistent with actual practice, in part because
actual central banks have thus far not adopted explicit loss functions.
In any event, the issue is of little importance, especially since it is
always possible to write instrument rules that approximate as closely
as desired the instrument settings of a policy regime involving targeting
in Svensson’s sense.”

11

See Benhabib, Schmitt-Grohé, and Uribe (2001), Svensson (2003), and
Svensson and Woodford (2003).

172

J U LY / A U G U S T 2 0 0 4

with the narrower definition of inflation targeting,
namely, the central role of macroeconomic forecasts
in inflation targeting, on the one hand, and the
robustness of the rules in view of the possible
presence of model misspecifications, on the other.

Information Content and Forecasts
As pointed out above, simple outcome-based
instrument rules constrain the central bank to
respond only to developments in observed inflation
and the output gap, and thus not to make use of other
available evidence about the state of the economy.
However, it is widely recognized that an efficient
monetary policy should exploit all relevant information. By imposing an arbitrary partition on the data,
simple instrument rules do not adopt such a fullinformation approach. This raises the issue of
whether those rules can be incentive-compatible.
If a central bank is aware of information suggesting
that the interest rate implied by the rule might be
inappropriate (e.g., because of weakness in the financial system), it would have an incentive to deviate
from the rule. Given the incentive for such deviations,
it is questionable whether central banks would follow
such a rule and thus whether the ex ante commitment to this rule can be credible. However, if the
rule lacked credibility, it is unlikely to help stabilize
private inflation and interest rate expectations.
Forecast-based rules partially overcome the
information restrictions imposed by outcome-based
Taylor-like rules, making the instrument respond
to expectations of future inflation and the output
gap. To quote Haldane and Batini (1998): “expected
inflation ought to embody all information contained
within the myriad indicators that affect the future
path of inflation.” Along the same lines, Clarida, Galí,
and Gertler (2000) characterize forecast-based rules
as making use of “a broad array of information
(beyond lagged inflation and output) to form beliefs
about the future condition of the economy, a feature
that we find highly realistic.”
However, this opens the door to the problem of
the complexity of the construction and nature of
the forecast. For example, which is the proper model
to forecast? What is the proper way of treating the
central bank’s monetary policy responses in the
future projection or of using market participants’
forecasts?12 Instead of tackling these issues, let me
12

See Bernanke and Woodford (1997) for an exposition of the circularity
problem induced by monetary policy mechanically responding to
private inflation forecasts.

FEDERAL R ESERVE BANK OF ST. LOUIS

focus on another main critique by challenging the
view that forecasts, particularly inflation forecasts,
are sufficient statistics on the state of the economy
and for monetary policy.
To exemplify, let us begin by assuming that the
only objective of policy is to maintain price stability.
If prices move in tandem with the existing tension
on employable resources, the policy goal of price
stability dictates keeping the economy continuously
close to its potential. Under these circumstances,
reacting to a pure inflation forecast figure, with no
reference to any additional indicator of macroeconomic performance, would be a recipe for policy
mismanagement any time the economy is hit by a
transitory (say, favorable) supply shock. This type
of shock would entail both a downward blip in forecast inflation and an upward movement of output
away from its sustainable level. Hence, restricting
the central bank’s information set solely to the inflation forecast figure—to the exclusion of a broader
suite of other indicators, which help discriminate
between supply and demand shocks—would, in
this situation, call for a policy easing today, which
could destabilize the economy by laying the ground
for the build-up of inflationary pressures tomorrow.
This is an elementary example of the general
proposition that inflation forecasts alone are not
sufficient to reveal the nature of the threat to price
stability and that it would therefore be misleading
to follow a rule that requires setting the policy instrument simply as the function of a forecast. Even in
an ideal world in which the models producing the
forecast are properly specified, the policymakers
are not interested in the result of the forecast per se
but instead aim at a consistent economic picture—
or, to put it differently, they aim at identifying the
relevant shocks underlying the forecasts and how
different types of disturbances to the economy imply
different kinds of policy responses. The relation
between forecasts and underlying shocks is clearly
one-to-one in many simple stylized models used in
the monetary policy literature, but this relation
clearly breaks down once we depart from that simple
set-up. So, once again, forecasts of a few macrovariables cannot be sufficient statistics to determine
monetary policy action.
Target rules are somehow immune from the
above problem, given that they are routinely implemented by producing forecasts of future inflation
and output conditional on the path of the policy
instrument and searching for the path, which minimizes a proper loss function. Consequently, when

Panel Discussion

evaluating inflation targeting in the context of target
rules, the discussion should primarily focus on the
structural model used to define the central bank’s
constrained optimization problem. In other words,
an evaluation of the target rule characterization of
inflation targeting is largely equivalent to an evaluation of the economic model employed to derive that
rule.13 One criticism of the models underlying most
target rule characterizations of inflation targeting is
that they neglect any role that might be played by
the monetary aggregate or financial frictions in the
determination of price developments. This opens the
way to a second set of remarks on the issue of model
misspecification and the robustness of the rules.

Robustness and Model Misspecification
The possible presence of model misspecification
is something that economists and econometricians
have some difficulty in acknowledging. However,
every model we write down and estimate contains
some form of shortcut and approximation. This
uncertainty is worsened since economists have not
yet agreed upon a proper, commonly accepted
approximating model. This implies that the appropriateness of a monetary policy strategy cannot be
evaluated only within a particular class of model—
rather a good strategy has to perform well across a
variety of empirically plausible models.
However, most advocates of inflation targeting—
at least those referring to simple rules for monetary
policy decisions—ultimately rely on a view of the
economy whose essence can be captured by no more
than three equations. The defining characteristics
of these equations are (i) staggered pricing, (ii) the
centrality of the output gap (or Phillips curve), and
(iii) the notion that monetary impulses propagate
primarily via a price (interest rate) channel, with
monetary quantities playing no role.
The presence of only a market for goods and
the absence of a fully formalized market for assets
whose supply is inelastic in the short run implies
that money has no role other than to facilitate the
exchange of goods. Decisions about money holdings
are not seen as part of a wider portfolio decision
that—at times—may lead households to prefer
liquidity over risky assets. For example, a positive
change in money demand has no counterpart in
13

While Svensson (2003, p. 450) says “[forecast targeting] does not imply
that forecasts must be exclusively model-based,” I would not tackle
the slippery issue of the use of judgmental information in the forecasting process.

J U LY / A U G U S T 2 0 0 4

173

REVIEW

Panel Discussion

an excess supply of some other asset. On the contrary, if truly alternative assets were to exist whose
issuance was related to their private issuers’ investment decisions and capital formation, then generating
a higher (or lower) supply of money—at any given
interest rate—could become all but inconsequential.
Quoting McCallum (2001a), there is “nothing
fundamentally misguided” about the model used
by advocates of inflation targeting. Such a model is
internally consistent and elegant. It can also mimic
the observed behavior of modern economies in
“normal” circumstances. Yet it rests upon what can
certainly be regarded as extreme assumptions about
the role of money in the economy. A central bank
can legitimately question the usefulness of a model
for monetary policy-setting in which money has
been deprived of its basic liquidity—or, equivalently,
its “store-of-value”—function that generations of
scholars have recognized and discussed for decades
(cf. Hahn, 1990, inter alia).
Levin, Wieland, and Williams (1999, 2001)
demonstrate that Taylor-like instrument rules perform quite robustly in a particular set of macroeconomic models. However, this robustness does
not survive a broadening of the suite of candidate
models beyond those considered in these papers.
Suitably parameterized Taylor-like rules appear to
work well in stabilizing the economy within the confines of the mainstream New Keynesian paradigm,
in which money market equilibrium conditions are
redundant. This last assumption, in particular, proves
to be absolutely crucial. If financial markets are not
free of frictions, then Taylor-like rules often do not
prove to be robust and yield suboptimal outcomes.
Examples of financial market frictions are prominent in transmission mechanism literature, e.g., in
so-called limited participation models (Christiano
and Eichenbaum, 1992) or segmented markets
models (Alvarez, Lucas, and Weber 2001). Within
the class of limited participation models, Christiano
and Gust (1999) show that the set of parameters
under which a Taylor-like inflation-targeting rule
becomes a source of instability is much broader
than for mainstream New Keynesian models.
More recently, Alvarez, Lucas, and Weber (2001)
presented some experiments on the stabilization
properties of simple Taylor rules within a segmented
financial markets model. They conclude that central
banks pursuing a Taylor-like interest rate instrument
rule—by systematically ignoring money market
(velocity) shocks—censor the information set available to policymakers and thereby reduce the effec174

J U LY / A U G U S T 2 0 0 4

tiveness of their responses to economic shocks by
arbitrarily excluding relevant monetary information
from the policy decision. In a similar vein, but in a
different class of model, Christiano and Rostagno
(2001) show that a Taylor-like interest rate instrument
rule can generate equilibria with undesirable properties; this outcome could be avoided by a policy
rule that takes into account the information provided
by monetary aggregates.
It should be clear that, from the view point of a
central bank, a serious attempt should be made to
construct a model where shocks to velocity are
treated appropriately within the context of broader
portfolio shifts, possibly in the presence of (changing)
risk assessments. Unless disturbances in money
holdings are formalized in such a way as to reflect
financial decisions, then nothing can be said about
the role of money in the business cycle and insufficient policy advice can be drawn from analyses of
models that do not properly tackle these problems.

4. THE ECB’S MONETARY POLICY
STRATEGY
Let me now turn to the ECB’s monetary policy
strategy. The ECB started to conduct policy in 1999
with the inception of the euro. While taking stock
of the experience of the central banks of the
Eurosystem, the ECB was at the time facing a major
institutional change. Eleven national economies14
merged into a unified market almost overnight; in
this context, past experience and data might turn
out not to be particularly informative with regard
to the new economic structure.
In the presence of such Knightian uncertainty,
in October 1998 the Governing Council of the ECB
announced its monetary policy strategy. The designed
strategy was a novel one, suited to the special and
still partially unknown characteristics of the euro
area, and different in a number of respects from
other current and past strategies.
Three aspects of the ECB strategy are critical.
First, its focus on the price stability objective: Price
stability is enshrined in the Treaty on European
Union as being the primary objective of the ECB. The
ECB Governing Council therefore provided a quantitative definition of price stability as a year-on-year
increase of the Harmonized Index of Consumer
Prices (HICP) below 2 percent.15 A second, closely
14

Twelve from January 2001, when Greece also adopted the euro.

15

See Issing et al. (2001) and ECB (2001a) for a detailed description of
the ECB’s monetary policy strategy.

FEDERAL R ESERVE BANK OF ST. LOUIS

Panel Discussion

related element, is the medium-term orientation of
our policy. Central banks can only affect the price
level with “long and uncertain lags”; consequently
they cannot be overambitious and try to steer price
developments in the short run, nor should they seek
to precisely define the horizon of their action. Moreover they need to respond gradually to economic
shocks, taking output fluctuations into account. A
third element of the strategy relates to the analyses
and economic perspectives that ultimately guide
policy decisions. The strategy recognizes the need
for a comprehensive analysis of economic and financial shocks and dynamics but, at the same time, it
attaches a privileged role to indicators based on
monetary aggregates. This organization of the information has been labeled the “two-pillar” strategy
of the ECB. The ECB’s monetary policy strategy was
meant to provide a transparent and consistent conceptual framework: structuring the internal analysis
of the economic situation and risks to price stability,
facilitating the decisionmaking process in the
Governing Council, and communicating policy
decisions to the public at large.
In almost five years of experience with the ECB
monetary policy, the strategy has served all these
functions to a high level of satisfaction. The ECB has
pursued its mandate of maintaining price stability
with vigor and determination, gaining a high level
of credibility from the outset. This achievement is
all the more remarkable given that the ECB started
without a track record and in an uncertain environment. Testifying to this success, inflation expectations, as measured by survey data and by financial
market indicators, have remained consistent with
our definition of price stability.16
In December 2002 the ECB announced the
decision to conduct a comprehensive review of its
strategy. This decision was sometimes wrongly interpreted by observers as an implicit indication of dissatisfaction with the strategy. In fact, the opposite
was true. To ensure the continued satisfactory development of the strategy in a complex and changing
environment, it was only natural that, after more
than four years of experience, the ECB Governing
Council would want to look back and reflect in a
systematic way on past experience. The outcome
of the strategy review was made public on May 8,
2003, and it aimed primarily at addressing certain

misunderstandings that have emerged in our communication with the public.17
Regarding the definition of price stability, the
Governing Council confirmed the explicit quantitative definition announced back in October 1998.
However, in continuing with the past conduct of
monetary policy, the Governing Council clarified
that in the pursuit of price stability it will aim to
maintain inflation rates below but close to 2 percent
over the medium term. This clarification emphasizes
the need for a sufficient safety margin against the
risk of deflation and at the same time is also sufficient
to cover the potential presence of a measurement
bias in the HICP and the implications of inflation
differentials of a structural nature within the euro
area.
Regarding the role of money in the strategy
framework, the Governing Council confirmed that
the strategy’s two-pillar framework is an effective
tool for organizing the information used to assess the
risks to price stability. As discussed in the previous
section, the economic literature confirms that integrating the analysis of monetary aggregates with
the analysis of conditions on the goods and labor
markets in a unified model remains an elusive
challenge. Different types of analysis provide information relevant for price developments at different
time horizons. What we labeled as economic analysis focuses on the most proximate causes of inflation,
such as cost developments and demand-supply
imbalances, and primarily contributes to the assessment of short- to medium-term economic dynamics
and the risks to price stability at that horizon. Monetary analysis, on the other hand, focuses instead on
the ultimate monetary determinants of inflation,
and primarily contains information for assessing
price trends at medium- to long-term horizons.
The Governing Council clarified that the monetary
analysis mainly serves as a means of cross-checking,
from a medium- to long-term perspective, the shortto medium-term indications coming from the economic analysis.
Let me emphasize the role of the cross-checking.
All the information coming from different sources,
such as short-term conjunctural indicators, quarterly
macroeconomic forecasts, the analysis of asset prices
and monetary aggregates, have to be compared and
properly evaluated to come to an overall assessment
of the monetary policy stance. This ensures that,

16

17

See the evidence provided in Castelnuovo, Nicoletti-Altimari, and
Rodriguez-Palenzuela (2003).

The outcome of the strategy review and the background documents
can be found at www.ecb.int.

J U LY / A U G U S T 2 0 0 4

175

REVIEW

Panel Discussion

while responding to economic shocks as they manifest themselves, we do not lose sight of the fact that,
in the longer term, developments in money need
to be consistent with our objective. This helps, in
my view, to give a sense of direction and impart a
steady course to the conduct of monetary policy.
The Eurosystem staff macroeconomic projections18 are one important input into the monetary
policy decision as a way of organizing a large amount
of information and helping to create a consistent
picture of possible future developments, but without making them the sole input for the conduct of
monetary policy. As discussed in the previous section, forecasts cannot be, per se, a sufficient statistic
for policy, nor can they contain all relevant information, not least because the models underlying the
forecast are inevitably misspecified to some extent.
There are instances that standard macroeconomic models, which, by definition, are constructed
to replicate normal conditions and regularities in
the economy, are unable to capture and incorporate.
This is particularly the case when large shocks or
special circumstances arise, such as episodes of
financial instability or asset price bubbles. I am
merely recalling the developments over the past two
to three years, when we faced exceptional uncertainties and major stock market movements followed
by large portfolio adjustment. How those past events
can be squared with forecasts of inflation and output, based on models in which financial assets do
not play any active role, is still an open issue both
for central bankers and academics. In such occasions
the need of careful judgment, of a broadening of
the horizon for the conduct of policy, and of the
consideration of non-standard indicators and different interpretations of the evidence become especially relevant.19
Of course this does not mean that the ECB does
not make full use of models. On the contrary, the
ECB devotes a lot of time and resources to improving
the set of economic models that are used in-house
to gain a better understanding of the euro area economy and provide better guidance for monetary
18

See ECB (2001b) for a description of the Eurosystem staff macroeconomic projection exercise. Within our framework, we clearly separate
the production of projections, as carried out under the responsibility
of the staff, from the monetary policy decisions taken under the responsibility of the Governing Council in order to avoid any ambiguity
between the assumptions of the projections and the policy implications.

19

See Issing (2002) for a discussion on the usefulness of information
stemming from monetary aggregates in revisiting some historical
episodes of financial instability.

176

J U LY / A U G U S T 2 0 0 4

policy decisionmaking. Like many other central
banks, we do use quite a large menu of models ranging from simple time-series models, useful in shortterm forecasting, up to medium-size structural
macroeconometric models in both area-wide and
multi-country specifications.20 Compared with
purely time-series or reduced-form models, structural
models have the advantage of having a well-specified
conceptual framework (or a set of identification
assumptions) that help to provide some better economic interpretation of the results, i.e., “the story
behind the numbers.” Moreover, considerable effort
has recently been devoted to the development of
“state of the art” medium-sized stochastic dynamic
general equilibrium models (SDGE), where the estimated specification is fully micro-funded and consistent with the solution of the optimization problem
of economic agents. Those models have proved to
combine a solid theoretical grounding with a good
ability to replicate many relevant features of the euro
area data. Smets and Wouters (2003) proposed an
extended version of the standard New Keynesian
SDGE closed-economy model with sticky prices
and wages. Christiano, Motto, and Rostagno (2003)
substantially extended a stylized real-sector SDGE
model to include a fully formalized financial sector,
where the issuance of assets is related to firms’ need
to finance entrepreneurial activity, although there
are frictions in the activity of the intermediaries
related to the cost paid to monitor firms.
As a central banker but also as an academic, I
am looking forward to the results of this line of
research given that it provides macro models with
both a solid micro-foundation and good empirical
properties, and with the potential to bring into the
picture phenomena of a monetary and financial
nature that are often left out of the more commonly
used macro models.

5. CONCLUSION
Let me end by saying that, in practice, probably
no central bank follows the strict characterization
of inflation targeting and that differences in the
practices of central banks oriented to price stability
should not be exaggerated. Most of the central banks
oriented to price stability share a number of key
elements that guide the conduct of their monetary
policy, namely, a clear, quantitative definition of the
overriding objective, a forward-looking orientation
20

See Fagan, Henry, and Mestre (2001) for a description of the area-wide
model of the euro area.

FEDERAL R ESERVE BANK OF ST. LOUIS

of their policy, and the awareness of the need to
take a broad range of information into account and
to communicate with the public in a clear and
transparent manner.
There is no clear-cut evidence to suggest that
generally, and according to some well-specified
criteria, one specific framework should be preferred
to all others.21 Take, for example, one crucial measure of our success as central bankers: the ability to
firmly anchor long-term inflation expectations.
These appear to be well anchored, in terms of the
very low volatility of expectations as well as the very
low correlation with actual inflation developments,
in most industrialized countries or currency areas,
including those where central banks do not usually
consider themselves to be “inflation targeters,” such
as the United States and the euro area. This to me
is just a confirmation of something that I always
believed: that there is no “single” or “best” way to
conduct monetary policymaking and that different
approaches or frameworks can lead to successful
policies and/or be better adapted to different institutional, economic, and social environments.

REFERENCES
Amato, Jeffrey D. and Gerlach, Stefan. “Inflation Targeting
in Emerging Market and Transition Economies: Lessons
After a Decade.” European Economic Review, April 2002,
46(4-5), pp. 781-90.
Alvarez, Fernando; Lucas, Robert E. Jr. and Weber, Warren
E. “Interest Rates and Inflation.” American Economic
Review, May 2001, 91(2), pp. 219-25.
Ball, Laurence and Sheridan, Niahm. “Does Inflation
Targeting Matter?” in Ben S. Bernanke and Michael
Woodford, eds., The Inflation Targeting Debate. Chicago:
University of Chicago Press (forthcoming).
Barro, Robert J. and Gordon, David B. “Rules, Discretion
and Reputation in a Model of Monetary Policy.” Journal
of Monetary Economics, 1983, 12(1), pp. 101-21.
Benhabib, Jess; Schmitt-Grohé, Stephanie and Uribe,
Martin. “Monetary Policy and Multiple Equilibria.”
American Economic Review, March 2001, 91(1), pp. 167-86.
21

See Ball and Sheridan (2003) for a comparative analysis of the macroeconomic performance of countries adopting an inflation-targeting
framework.

Panel Discussion

Bernanke, Ben S.; Laubach, Thomas; Mishkin, Frederic S.
and Posen, Adam S. Inflation Targeting: Lessons from the
International Experience. Princeton: Princeton University
Press, 1999.
Bernanke, Ben S. and Gertler, Mark. “Monetary Policy and
Asset Price Volatility.” NBER Working Paper No. 7559,
National Bureau of Economic Research, February 2000.
Bernanke, Ben S. and Mishkin, Frederic S. “Inflation
Targeting: A New Framework for Monetary Policy?”
Journal of Economic Perspectives, Spring 1997, 11(2), pp.
97-116.
Bernanke, Ben S. and Woodford, Michael. “Inflation
Forecasts and Monetary Policy.” Journal of Money, Credit,
and Banking, November 1997, 24(4, Part 2), pp. 653-84.
Castelnuovo, Efrem; Nicoletti-Altimari, Sergio and
Rodriguez-Palenzuela, Diego. “Definition of Price Stability,
Range and Point Inflation Targets: The Anchoring of
Long-Term Inflation Expectations.” ECB Working Paper
No. 273, European Central Bank, 2003.
Clarida, Richard; Galí, Jordi and Gertler, Mark. “Monetary
Policy Rules and Macroeconomic Stability: Evidence and
Some Theory.” Quarterly Journal of Economics, February
2000, 115(1), pp. 147-80.
Christiano, Lawrence J. and Eichenbaum, Martin. “Liquidity
Effects and the Monetary Transmission Mechanism.”
American Economic Review, May 1992, 82(2), pp. 346-53.
Christiano, Lawrence J. and Gust, Christopher J. Comment
on A. Levin, V. Wieland, and J.C. Williams, “Robustness of
Simple Monetary Policy Rules under Model Uncertainty,”
in J.B. Taylor, ed., Monetary Policy Rules. Chicago:
University of Chicago Press, 1999.
Christiano, Lawrence and Rostagno, Massimo. “Money
Growth Monitoring and the Taylor Rule.” NBER Working
Paper No. 8539, National Bureau of Economic Research,
October 2001.
Christiano Lawrence; Motto, Roberto and Rostagno,
Massimo. “A Monetary Dynamic General Equilibrium
Model of the Euro Area.” Unpublished manuscript,
European Central Bank, 2003.
European Central Bank. The Monetary Policy of the ECB.
Frankfurt: European Central Bank, 2001a; www.ecb.int.

J U LY / A U G U S T 2 0 0 4

177

Panel Discussion

European Central Bank. A Guide to Eurosystem Staff
Macroeconomic Projection Exercises. Frankfurt: European
Central Bank, 2001b; www.ecb.int.
Fagan, Gabriel; Henry, Jerome and Mestre, Ricardo. “An
Area-Wide Model (AWM) for the Euro Area.” ECB
Working Paper No. 42, European Central Bank, 2001.
Haldane, Andrew G. and Batini, Nicoletta. “Forward-Looking
Rules for Monetary Policy.” NBER Working Paper No.
6543, National Bureau of Economic Research, 1998.
Hahn, Frank H. “Liquidity,” in B.M. Friedman and F.H. Hahn,
eds., Handbook of Monetary Economics. Amsterdam:
North Holland, 1990.
Issing, Otmar. “Why Price Stability?” in Herrero et al., eds.,
First ECB Central Bank Conference: Why Price Stability?
Frankfurt: European Central Bank, 2001, pp. 179-202.
Issing, Otmar. “Central Bank Perspectives on Stabilization
Policy.” Federal Reserve Bank of Kansas City Economic
Review, Fourth Quarter 2002, 87(4), pp. 15-36.
Issing, Otmar; Gaspar, Vitor; Angeloni, Ignazio and Tristani,
Oreste. Monetary Policy in the Euro Area: Strategy and
Decision-Making at the European Central Bank. Cambridge:
Cambridge University Press, 2001.
Kydland, Finn E. and Prescott, Edward C. “Rules Rather
Than Discretion: The Inconsistency of Optimal Plans.”
Journal of Political Economy, June 1977, 85(3), pp. 473-91.
Kohn, Donald L. Comment on M. Goodfriend, “Inflation
Targeting in the United States,” in Ben S. Bernanke and
Michael Woodford, eds., The Inflation Targeting Debate.
Chicago: University of Chicago Press (forthcoming).
Levin, Andrew; Wieland, Volker and Williams, John C.
“Robustness of Simple Monetary Policy Rules under
Model Uncertainty,” in John B. Taylor, ed., Monetary
Policy Rules. Chicago: University of Chicago Press, 1999.
Levin, Andrew; Wieland, Volker and Williams, John C. “The
Performance of Forecast-Based Rules for Monetary
Policy Rules under Model Uncertainty.” ECB Working
Paper No. 68, European Central Bank, July 2001.
Lucas, Robert E. Jr. “Econometric Policy Evaluation: A
Critique,” in K. Brunner and A.H. Meltzer, eds., The
Philips Curve and Labor Markets. Carnegie-Rochester
178

J U LY / A U G U S T 2 0 0 4

REVIEW
Conference Series on Public Policy. Volume 1. Amsterdam:
North Holland, 1976, pp. 19-46.
Marschak, Jacob. “Economic Structure, Path, Policy and
Prediction.” American Economic Review, May 1947,
37(2), pp. 81-84.
McCallum, Bennett T. “Monetary Policy Analysis in Models
Without Money.” NBER Working Paper No. 8174,
National Bureau of Economic Research, March 2001a.
McCallum, Bennett T. “Inflation Targeting and the Liquidity
Trap.” NBER Working Paper No. 8225, National Bureau
of Economic Research, April 2001b.
Orphanides, Athanasios and Williams, John C. “The Decline
of Activist Stabilization Policy: Natural Rate Misperceptions,
Learning and Expectations.” Unpublished manuscript,
Board of Governors of the Federal Reserve System, 2003.
Rodriguez-Palenzuela, Diego; Camba-Méndez, Gonzalo and
Garcia, Juan A. “Relevant Economic Issues Concerning
the Optimal Rate of Inflation.” ECB Working Paper No.
278, European Central Bank, 2003.
Simons, Henry C. “Rules versus Authorities in Monetary
Policy.” Journal of Political Economy, February 1936,
44(1), pp. 1-30.
Sims, Christopher A. “Limits of Inflation Targeting,” in Ben
S. Bernanke and Michael Woodford, eds., The Inflation
Targeting Debate. Chicago: University of Chicago Press
(forthcoming).
Smets, Frank and Wouters, Rafael. “An Estimated
Stochastic Dynamic General Equilibrium Model of the
Euro Area.” Journal of European Economic Association,
September 2003, 1(5), pp. 1176-206.
Svensson, Lars E.O. “Inflation Forecast Targeting:
Implementing and Monitoring Inflation Targets.” European
Economic Review, June 1997, 41(6), pp. 1111-46.
Svensson, Lars E.O. “Inflation Targeting as a Monetary
Policy Rule.” Journal of Monetary Economics, June 1999,
43(3), pp. 607-54.
Svensson, Lars E.O. “How Should Monetary Policy Be
Conducted in an Era of Price Stability?” NBER Working
Paper No. 7516, National Bureau of Economic Research,
2000.

FEDERAL R ESERVE BANK OF ST. LOUIS

Panel Discussion

Svensson, Lars E.O. “Inflation Targeting: Should It Be
Modeled as an Instrument Rule or a Targeting Rule?”
European Economic Review, April 2002, 46(4/5), pp. 771-80.

Taylor, John B. “Discretion versus Policy Rules in Practice.”
Carnegie-Rochester Conference Series on Public Policy,
1993, 39, pp. 195-214.

Svensson, Lars E.O. “What Is Wrong with Taylor Rules?
Using Judgment in Monetary Policy through Targeting
Rules.” Journal of Economic Literature, 2003, 41(2), pp.
426-77.

Walsh, Carl E. Monetary Theory and Policy. Cambridge, MA:
MIT Press, 1998.

Svensson, Lars E.O. and Woodford, Michael. “Implementing
Optimal Policy through Inflation-Forecast Targeting.”
Unpublished manuscript, Princeton University, 2003.

Inflation Targeting
Donald L. Kohn

I

should start with two declarations. First, the
usual disclaimer holds with particular force
today—the views I am about to express are
my own and not necessarily those of any other
policymaker at the Federal Reserve. Second, this
conference has been most interesting and informative, but I remain an inflation targeting (IT) skeptic.
I will briefly lay out the reasons for my attitude,
then address some topics, such as communication,
that frequently arise in the discussion, and conclude
by trying to stress-test my skepticism by speculating
on whether IT would have been helpful in some
recent episodes related to monetary policy.

INFLATION TARGETING FOR THE
UNITED STATES
I agree with advocates of IT in several critical
areas. Price stability—or its approximation at very
low inflation—is the appropriate primary long-term
objective of monetary policy, and achieving this
objective is the way that policy can best contribute
to the long-term welfare of the country. Moreover,
in some countries, adopting IT, together with the
central bank independence that often accompanies
the initiation of IT regimes, has been a major step
toward attaining price stability.

The question I would like to address is whether
IT would improve economic performance in the
United States. That is, would IT be likely to lead to
actions by policymakers and private agents that
increase the odds on keeping the economy producing at its maximum sustainable level and inflation
low and stable. In my view, the verdict on IT for the
United States is at least “not proven” and possibly
negative—that is, IT might detract from economic
performance over time.
I start from the premise that the United States
has had a very successful monetary policy over the
past two decades. We have achieved price stability,
inflation expectations are low and stable, and we
have done this with two relatively shallow recessions
in 20 years. Many factors have contributed to this
economic performance, but monetary policy has
been an important element. So, for me, the default
option is to keep doing what we have been doing—
however hard it might be to model or explain. And
that is not inflation targeting. I believe that adopting
IT, even in its softer versions, would be a slight shift
along the continuum of constrained discretion in
the direction of constraint, and the benefits of
such a shift are unlikely to outweigh its costs. Consequently, I would stick with the status quo.
On the cost side, I believe that under some
circumstances central banks do face short-term
trade-offs between economic stability and inflation
stability, and I am concerned that IT would result in
less-than-optimal attention being paid to stabilizing
the economy and financial markets. In its actions,

Donald L. Kohn is a member of the Board of Governors of the Federal Reserve System.
Federal Reserve Bank of St. Louis Review, July/August 2004, 86(4), pp. 179-83.
© 2004, The Federal Reserve Bank of St. Louis.

J U LY / A U G U S T 2 0 0 4

179

Panel Discussion

the Federal Reserve has put considerable weight
on achieving and maintaining price stability, but it
has not been “inflation targeting”—not even implicitly. IT implies putting a higher priority on hitting a
particular inflation objective over the intermediate
run than the Federal Reserve has done.
This point is most obvious from 1983 to 1997,
in the so-called opportunistic disinflation period.
During this time, the Federal Reserve was well aware
that inflation was running above levels consistent
with price stability but concentrated on keeping
inflation from rising, not on reducing it further.
I believe the Federal Reserve also paid more
attention to non-inflation factors than IT would have
suggested in the 1997 to 2003 period, even though
inflation outcomes were low and stable. Its broader
focus was especially evident in the reaction to the
threat to financial stability in the fall of 1998 and
in the very aggressive easing in early 2001. In the
latter case, easing continued through the spring
even though inflation expectations looked as though
they might be increasing, which would have been
very difficult for an IT central bank to ignore. I recognize that such responses would in theory be available under flexible IT, but I wonder what would
happen in practice. Most IT frameworks put a priority
on inflation control and base their communication
and accountability structures on inflation forecasts
and outcomes. Under circumstances in which shortrun conflicts among various objectives are possible,
I ask myself where IT policymakers are likely to
take their chances.
Moreover, with its concentration on mean
inflation, IT seems to be ill-adapted to the riskmanagement paradigm that Chairman Greenspan
(2003) laid out in Jackson Hole. That mode of operation, which I believe has been an important factor
in the Federal Reserve’s success, weighs the skews
in the outlook, as well as the central tendencies, and
also takes account of the cost of missing on one side
or the other—and for more than one objective.
I think that the U.S. economy has benefited
from the flexibility that the Federal Reserve has
derived by eschewing a formal inflation target. By
flexibility I mean not frequent changes in long-term
objectives but rather the freedom to deviate from
long-term price stability, perhaps for a while. I recognize that such deviations are also possible in
models of flexible IT, but I question whether they
can occur in practice.
180

J U LY / A U G U S T 2 0 0 4

REVIEW
Against these potential costs, I believe that the
benefits of IT in the United States relative to the
current regime are questionable.
We do not see evidence in IT economies that
inflation is lower or more stable or that output is
more stable around potential. On the surface, then,
IT appears to produce little or no gain in hitting goals.
To be sure, the evidence on how well inflation expectations are anchored is more mixed. Levin, Natalucci,
and Piger (2004) at this conference, provided some
backing for the idea that long-term expectations in
IT economies respond less to incoming information
on inflation. But I am also aware that the bulk of
the studies show that interest rates and inflation
are no more predictable in IT economies than in
non-IT economies. The IT economies examined in
the studies may have been subject to larger shocks
than the non-IT economies studied, but the burden
of proof should be on the advocates of IT to show
that it would improve economic performance in nonIT economies—by providing either greater cyclical
stability or better resource allocation.
A frequently used argument for IT in the United
States is that it will help to extend the good performance of monetary policy as leadership changes—
that is, it will protect against persistent increases or
decreases in inflation under a new chairman. In
my view, however, considerable safeguards against
these outcomes are already in place. The law mandates price stability. Without exception, everyone
now on the Federal Open Market Committee (FOMC)
agrees with this mandate, and it enjoys wide acceptance in the public and in the Congress as well as in
the academic community. Moreover, FOMC members
have diverse views and the Committee has been
operating in an environment in which members
are free to express those views—in sharp contrast
to some earlier eras. Any chairman gets deference,
but a new chairman would not have the clout of
Alan Greenspan, at least initially. A further safeguard
is provided by the greater amount of public discussion and media attention to monetary policy currently than in the 1960s and 1970s.
Of course there is a risk, however small, that
incompetence or political motivations in a new
leader might foster new trends or greater variability
in inflation, and IT might help counter any such
tendencies. The question is whether insuring against
this remote outcome is worth paying the cost. IT
prevents some bad results, but it tends to foreclose
very good results as well.

FEDERAL R ESERVE BANK OF ST. LOUIS

SPECIAL TOPICS
Communications and Transparency
IT does provide a clear framework for communicating with the public if communication is framed
mostly around the behavior of inflation relative to
the target. But does it help produce better policy
and economic outcomes? For flexible inflation targeters who are paying attention to other objectives
as well as inflation, communication tends to be clear
but not especially transparent. Other goals are
downgraded. In practice, IT communication does
not even mention varying time periods for achieving
price stability, much less the reason for those periods
to vary. Those other goals are the tough messy stuff
that does not fit into the IT framework very well.
That they get so little attention is not surprising
because accountability is usually framed in terms
of inflation and the reports are elements in the
accountability framework. But if, in fact, the goals
of economic and financial stability are factored into
policy decisions, they are often poorly acknowledged
in IT communication. There is also a risk that communication will drive policy, and so those goals end
up with less-than-optimal attention. For the most
part, the manifestations of better transparency—
reduced variability and greater predictability of
inflation and interest rates—are not readily apparent
in IT economies.
I am not arguing that the Federal Reserve cannot communicate better. But I am saying that IT is
not a cure-all for communication problems; that it
might not even help much in the markets where it
really counts; and that, if simplicity of communications drives policy, IT might lead to inferior economic outcomes.

Political Legitimacy
In his paper at this conference, Larry Meyer
(2004) was right to emphasize the importance of
the Federal Reserve’s interactions with the political
system. One of the major values of IT is its role in
forcing the people and their representatives to think
through carefully what they can and cannot expect
or demand from a central bank. This benefit would
be lost through unilateral adoption of IT by the
Federal Reserve.
The Federal Reserve is in a more complex position within the government relative to the central
banks of many other countries, and this position
both complicates any consultative process and ele-

Panel Discussion

vates its importance. The checks and balances of
our system mean that, unlike most other central
banks, which operate in a parliamentary system, we
do not have a “government” to interact with. The
paradigm of goal dependence/instrument independence so common in IT regimes is effectively blocked
for us. If we moved toward setting a goal for ourselves, perhaps even if we just defined price stability,
we would need to consult carefully with both houses
of the Congress and the Administration and would
need to judge what, short of legislation, constituted
a veto by any of the people with whom we were
consulting. This process would be subtle and difficult—but absolutely essential to protect our independence and preserve our democratic legitimacy.

Defining Price Stability
By “defining price stability,” I mean publishing
a number or a reference range that makes more
concrete our long-term inflation objective, without
making a commitment to achieve that objective in
any given time frame. Individual FOMC members
are increasingly stating their numerical definition
of price stability, but the Federal Open Market Committee has not done it. In some respects, such a
specification is an appealing idea. In concept, it
might allow the United States to realize some of
the benefits of inflation targeting without some of
the costs. The theory would be that putting a numerical value on long-term price stability could reduce
uncertainty about longer-run price tendencies without constraining our actions to stabilize the economy or financial markets over shorter periods.
I am still trying to make up my mind on the
balance of costs and benefits of taking this step. As
I have already noted, most evidence does not suggest
a lot of private uncertainty about longer-term price
trends in the United States, and so the benefits, if
any, would be limited. Spreads between nominal
and indexed 10-year bonds have fluctuated narrowly
around 2 percent since 1999, and survey measures
of long-term inflation expectations have barely
moved in recent years. Nonetheless, further evidence supporting the inference of Levin, Natalucci,
and Piger that IT would result in even more firmly
anchored expectations would be important in this
equation.
The costs, given my views on IT, would arise
from any tendency for this definition to morph into
a target that unnecessarily constrained actions—
that did not effectively permit outcomes outside the
range or away from the target under some circumJ U LY / A U G U S T 2 0 0 4

181

Panel Discussion

stances. Resisting such a tendency would be difficult,
I think, once the number was given. And the pressure to elevate price stability over economic stability,
even in the short-term, would be accentuated
because the latter goal would not have a numerical
value. However, ways of mitigating this tendency
might be found—for example, by giving a fairly
wide range and making it clear that the midpoint
had no special meaning and that the edges were
soft. Critical to maintaining useful flexibility would
be the understanding, believability, and sustainability of the “provisos” that the Federal Reserve would
give outlining the circumstances under which it
would not seek to achieve its price stability objective.

STRESS-TESTING MY SKEPTICISM—
WOULD INFLATION TARGETING
HAVE IMPROVED ECONOMIC
PERFORMANCE IN RECENT YEARS?
1. Would IT have contributed in any way to
damping the boom and bust since the mid-1990s?
I have already voiced my opinion that it would not
have helped and might even have hurt in the reaction to emerging weakness in 2001. But another
part of the question is, Would IT have constrained
the previous upswing in a way that also would have
lessened the subsequent weakening?
A number of observers believe that a little more
policy tightening a little earlier might have damped
the fluctuations in financial markets and the economy. Personally, I doubt that, given the strong forces
at work. But I also do not think an IT framework
would have helped, even if such an outcome were
possible. Inflation was edging lower through much
of this period. To be sure, forecasts were consistently
missing on the high side, so a forecast-based IT
framework might have run a slightly tighter policy—
but I do not think you want to rest a case for changing policy regimes on persistent forecast misses.
The arguments usually given for tighter Federal
Reserve policy in the mid-to-late 1990s reference
developments in asset prices—specifically in the
equity market—and the judgment that too-low interest rates fostered an intertemporal misallocation of
resources in the form of an excessive buildup of
capital and, hence, raised the amplitude of longerterm economic fluctuations. IT is especially poorly
adapted to deal with these sorts of issues, however,
since it tends to emphasize the performance of
inflation in consumer goods and services over the
succeeding few years. For those, like me, who are
182

J U LY / A U G U S T 2 0 0 4

REVIEW
skeptical about the ability of central banks to deal
with swings in asset prices or with longer-term
resource allocation issues, this aspect of IT is not
negative. Nonetheless, it is also evident in speeches
and commentary that policymakers in IT countries
right now are wrestling with the tension between
IT frameworks and the suspicion that economic
imbalances and disequilibriums in house or other
asset prices are developing that could disrupt the
economy at some point down the road.
2. Would ongoing IT or even a numerical definition of price stability have damped the bond market
volatility of this spring and summer?
Long-term interest rates fell steeply in May and
early June and rebounded even more sharply in late
June and July. The decline got under way in earnest
after the FOMC statement of May 6. What was the
news that day? First, the FOMC thought that inflation
could be below a level consistent with satisfactory
economic performance over time and that the current rate of inflation was close to that excessively
low level. Second, the FOMC was worried that the
lower limit would be breached—it thought inflation
was more likely headed down than headed up from
the already low level. In response, 10-year Treasury
rates fell 8 basis points the day of the announcement
and another 12 basis points the next day as the
import of the announcement sank in.
In my view, most of this immediate 20-basispoint decline in longer-term rates came not in
response to the clarification of the inflation objective but rather to the revelation that the FOMC was
worried about the trend of inflation. Moreover, much
of the information on the latter point was quite
recent, reflecting what seemed to be a lack of a
rebound in the economy after the Iraq war and a
steep decline in recent inflation readings. In these
circumstances, had we had an inflation target for a
while, rates might have been lower before the
announcement, but most of that decline would have
filtered into the markets only over the preceding
few weeks and rates would have been just as low a
few days after the meeting.
The next 70 basis points of rate decline occurred
by mid-June in response to further indications of
weakness in activity and prices and to statements
by Federal Reserve officials that they were thinking
about how to conduct policy in the remote contingency that a deflation threatened to take hold. I do
not see how this response would have been different in an IT framework. My judgment in this regard
is reinforced by the fact that rates in many IT coun-

FEDERAL R ESERVE BANK OF ST. LOUIS

tries over this same period of May and June fell by
a similar magnitude. Weakness in the world’s most
important economy and declines in its exchange
rate should lead rates overseas to decline, but the
extent and similarity of the decline is surprising.
This occurrence has led me to conclude that the
rate drop in the United States was caused by the
downward shocks to expected prices and activity,
not by the policy framework.
The IT countries did experience somewhat
smaller rate increases relative to the United States
in July and August. They did not have some of the
special factors pushing U.S. rates up—revised expectations about bond purchases and mortgage hedging
activity. Perhaps more importantly, their economies,
though strengthening, did not demonstrate the surprising degree of rebound that seems to be occurring
in the United States.
In sum, this is a striking episode in which misunderstandings between the central bank and the
markets probably contributed to an extraordinary
volatility in financial markets. But these misunderstandings did not stem from the absence of inflation
targets in the United States; volatility would have
been damped only a little, if at all, under IT.

CONCLUSION
I recognize that I am at risk of being interpreted
as saying that something good—the policy regime
of the past 20 years—cannot be made better or that

Panel Discussion

there are not downside risks to highly judgmental,
flexible policy with an imprecise price stability
objective. That is not what I think. I am open to
alternatives that promise improvements or that
raise the odds on good policy continuing in the
future without incurring much in the way of current
costs. But I do believe that those who propose
changes from a good system have a high burden of
proof. The marginal benefits from improving a good
regime, by definition, are not likely to be high. And
any change must deal with the uncertainties created
by the law of unintended consequences. I have yet
to be convinced that for the United States inflation
targeting has jumped those hurdles.

REFERENCES
Greenspan, Alan. “Monetary Policy under Uncertainty.”
Remarks given at a symposium sponsored by the Federal
Reserve Bank of Kansas City, Jackson Hole, Wyoming,
August 29, 2003.
Levin, Andrew; Natalucci, Fabio M. 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.
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.

J U LY / A U G U S T 2 0 0 4

183

REVIEW

184

J U LY / A U G U S T 2 0 0 4