Download Optimal Forward Guidance in Monetary Policy: Can Central Banks

Optimal Forward Guidance in Monetary Policy:
Can Central Banks Sway the Public with
Projections?
Christian Jensen∗
University of South Carolina
February, 2017
Abstract
Optimal central bank forward guidance through publicly released economic
projections requires misleading the public. Optimal non-misleading projections
are time-inconsistent. Non-misleading time-consistent projections only improve
policy outcomes if the public’s expectations are inconsistent with actual policy,
or less precise than policymakers’ forecasts. Since the public only has incentives to believe policymakers’ projections when they are most vulnerable to
being mislead, they cannot trust central bank projections, making these all but
irrelevant. The systematic deviations observed between FOMC projections and
survey forecasts indicate that the public is not fully convinced by these.
JEL Codes: E52; E58
Keywords: Central bank projections; forward guidance; survey forecasts;
managing expectations; monetary policy; commitment
∗
Department of Economics; University of South Carolina; 1705 College Street; Columbia, SC
29208; 803-777-2786; [email protected].
1
Introduction
In recent years many central banks, including the Federal Reserve, have started
releasing projections of future values of macroeconomic variables such as GDP
growth, inflation, unemployment, and even forecasts of future interest-rate targets.1 If policymakers’ projections influence the public’s expectations, they can
affect policy outcomes, since current inflation and output are posed to depend
on expected future inflation, according to the popular New-Keynesian Phillipscurve model.2 More generally, there appears to be a growing consensus that
monetary policy affects the economy mainly through private-sector expectations.3 Each central bank does things differently, and it is beyond our scope
to describe all these practices in detail. Instead, we study what role publishing
projections can play in policymaking. We find that optimal projections are
misleading, with the most desirable outcomes arising when the public’s expectations are distorted away from the values consistent with implemented policy.
Optimal non-misleading projections are time-inconsistent, just as optimal commitment. Optimal non-misleading time-consistent projections have no impact
on policymaking, assuming rational expectations, as they would not alter the
public’s expectations. Without rational expectations, projections can improve
policy outcomes even when non-misleading and time-consistent, if the public’s
forecasts are noisier than policymakers’, or if they are inconsistent with actual
policy. Hence, the public only has incentives to believe central bank projections
when they are most vulnerable to being mislead. It is therefore unclear whether
such projections will have any influence on the public’s expectations.
It is well known from the rules versus discretion literature (Kydland and
Prescott (1977) and Barro and Gordon (1983a, b)) that policymakers can
achieve their objectives to a greater extent if they can control expectations
by credibly committing to implement a rule in all periods. Releasing projections provides an alternative way to influence the public’s expectations, and
thus improve policy outcomes. Being more direct, projections make it easier
for the public to arrive at the desired expectations, as these do not have to
1
Projections are also provided by the ECB, and central banks of Australia, Canada, England,
Japan, New Zealand, Norway, Sweden and Switzerland.
2
While internal projections have played a role in monetary policy for a long time, for example
through forecast targeting (Hall (1985), Hall and Mankiw (1990), King (1994), Bowen (1995) and
Svensson (1997 and 1999)), our focus is on the more recent practice of making projections publicly
available (Svensson (2005) and Woodford (2007)).
3
See for example Woodford (2004), Svensson (2005), and Blinder et al. (2008).
2
be derived from a rule. But, projections are typically less informative, just a
value, or set of values, instead of a rule, making it easier to deceive the public.
For example, it is easier to verify that 5% inflation misses the established 2-3%
target, than determining if a realized rate of 5% means that a previously projection of 3% was biased or misleading. The latter is particularly difficult when
no details are provided about how the projections are generated, so that any
deviations from the projected path can be blamed on unanticipated shocks. Because of this, and the inherent flexibility in case of structural change, a central
bank might prefer to influence expectations through projections, rather than
through a once-and-for-all commitment to a rule.
Our point is not that the Fed, or any other central bank, uses projections
to mislead the public, but rather that they easily could, and have incentives to
do so, which is consistent with previous findings that policymakers often have
incentives not to reveal private information truthfully (Cukierman and Meltzer
(1986)). Consequently, the public should not believe central bank projections,
or at least, should not believe these blindly. Of course, if the public is not
influenced by the projections, there may be no gain, in terms of economic
outcomes, from providing these. In the absence of a perfect forecasting record,
convincing the public to believe the projections may require providing enough
information to enable everyone to ensure that they are not being mislead. But,
then the projections can only improve outcomes if the public’s expectations
were previously inconsistent, or more noisy. Releasing projections could even
be damaging, for example by correcting the public’s inconsistent expectations,
or if they induce people to believe that policymakers are less willing, or less
able, to fight inflation, leading to higher inflation expectations.
For simplicity, we assume in our analysis that the public either fully believes policymakers’ projections, or is completely unaffected by these. Reality is probably somewhere in between these two extremes, where projections
have some influence on people’s expectations, but do not allow policymakers
to control these perfectly. However, deriving the optimal projections for such
a case would require specifying exactly how these affect expectations, which
is unknown. Rather than making arbitrary assumptions, we use the two extremes to learn about optimal projections, finding that projections are likely
to have little, if any, impact on people’s expectations. This stands in stark
contrast with existing studies that assume the public believes policymakers’
projections blindly, so that their expectations match these exactly, even when
time-inconsistent (Svensson (2005) and Woodford (2007)). The empirical part
3
of our study also rejects such an unrealistic assumption.
We study the Federal Open Market Committee’s (FOMC) projections of
real GDP growth, unemployment and output. The Fed was chosen due to the
availability of comparable forecasts, mainly the Survey of Professional Forecasters (SPF). In addition, the available meeting transcripts provide an insight
into the FOMC’s deliberations. FOMC projections of output growth do not
appear to differ systematically from survey forecasts over the years studied,
2008-15, though the comparison data is very limited. While FOMC projections
for unemployment are systematically lower than in the surveys for 2013-15, the
differences are not statistically significant for the 2008-15 period as a whole.
There is highly statistically significant evidence for FOMC inflation projections
consistently being lower than those of the SPF from 2011 to 2015, and for the
2008-15 period as a whole. The Fed does not appear to be trying to mislead the
public, as FOMC projections are often more accurate than the SPF’s. However, despite its good record, the FOMC’s projections have failed to dictate the
public’s inflation expectations.
The next section provides a brief introduction to the New-Keynesian stickyprice model, which is widely used to study, and guide, monetary policy. The
following section shows that optimal projections are misleading, and can lead
to better economic outcomes than the standard optimal commitment policy. In
addition, it shows that providing misleading projections can improve outcomes
even after these are no longer believed by the public. The subsequent section
shows that the optimal non-misleading projections are time-inconsistent in the
same way, and for the same reason, that optimal plans are. The following
section studies optimal non-misleading time-consistent projections, which can
only improve policymaking if the public’s expectations are inconsistent with
the implemented policy, or more noisy than policymakers’ projections. The
subsequent section studies the FOMC’s experience with releasing projections.
2
Model
In a framework with monopolistic competition, menu costs and staggered pricesetting, Rotemberg (1982) and Calvo (1983) derive the New-Keynesian Phillips
curve
πt = β Êt πt+1 + κxt + ut
4
(1)
which relates present inflation πt to expected next-period inflation Êt πt+1 , contemporaneous output xt and a cost-push shock ut . It is the public’s (pricesetters’) inflation expectations that are relevant in the Phillips curve, and these
may or may not be rational, depending on the case below, so we denote these
as Êt to distinguish from Et , which is usually understood to imply rational
expectations. The exogenous shock is assumed to be known to satisfy
ut = ρut−1 + at
(2)
where at is white noise. The parameters β ∈ (0, 1) , κ > 0 and ρ ∈ (0, 1)
are assumed to be known constants. Inflation πt and output xt are measured
in terms of deviations from their flexible-price values, that is, the values they
would attain in the absence of menu costs. Hence, as Rotemberg and Woodford
(1999) and Woodford (2003) show, minimizing the distortions that menu costs
generate, by limiting the impact on the utility of a representative consumer,
can be accomplished at any time t = 0 by minimizing
L0 = E0
∞
X
β t lt
(3)
t=0
where
lt = πt2 + λx2t
(4)
and λ > 0 is a known constant.4 Details and extensions of this popular framework are discussed by Clarida et al. (1999), Svensson and Woodford (2002),
and Woodford (2003), among others.
From the perspective of any period t = 0, the optimal commitment plan is
the infinite sequence of current and future inflation {πt }∞
t=0 that minimizes the
loss function (3) subject to the Phillips curve (1). As shown by Clarida et al.
(1999) and Woodford (1999), it can be found by minimizing
£=
∞
X
β t πt2 + λx2t − θt (βπt+1 + κxt + ut − πt )
(5)
t=0
assuming rational expectations, and exploiting certainty equivalence (Currie
and Levine (1993)). Here, the variable θt is a Lagrange multiplier associated
4
We assume there is no role for interest-rate smoothing, and no bias in policymakers’ objectives.
5
with the Phillips-curve constraint (1). The first-order conditions,
∂£
= 2π0 + θ0 = 0,
∂π0
(6)
∂£
= 2β t πt − βθt−1 + θt = 0
∂πt
(7)
∂£
= 2β t λxt − κθt = 0
∂xt
(8)
for t = 1, 2, 3, . . . and
for t = 0, 1, 2, . . . , yield the optimal commitment policy
and
λ
π0 = − x0
κ
(9)
λ
λ
πt = − xt + xt−1
κ
κ
(10)
for t = 1, 2, 3, . . . While this optimal plan minimizes the objective function
(3), it is, as argued by Kydland and Prescott (1977), not time-consistent. The
reason is that if policymakers reoptimized in any later period τ > 0, they
would want to deviate from the promised commitment rule (10) in period τ
and instead enforce
λ
π τ = − xτ
κ
(11)
λ
λ
πt = − xt + xt−1
κ
κ
(12)
while committing to implement
in all later periods t = τ + 1, τ + 2, τ + 3, . . . If policymakers reoptimized every
period, they would implement
λ
πt = − xt
κ
(13)
in all periods t = 0, 1, 2, . . . making it the optimal discretionary policy. The
policy objective (3) is strictly positive under both optimal commitment and
discretion. Exactly how much lower it is with commitment depends on the
parameters, which there is disagreement about, particularly with respect to κ
and λ.5
5
See Assenmacher-Wesche (2006), Dennis (2004), Givens (2012), Schorfheide (2008) and
Söderström et al. (2005)
6
3
Optimal misleading projections
Imagine policymakers implement the policy
πt = 0
(14)
in all periods t = 0, 1, 2, . . . , while at the same time projecting future inflation
using
π̂t = −β −1 ρ−1 ut
(15)
for t = 1, 2, 3, . . . If the public believes these projections, despite being inconsistent with the policy actually implemented (14) whenever ut 6= 0,
Êt πt+1 = Et π̂t+1 = −β −1 ut
(16)
for t = 0, 1, 2, . . . , given the knowledge of the exogenous shock process (2). Inserted into the Phillips curve (1), together with the implemented policy (14), it
implies yt = 0 and lt = 0 for all t = 0, 1, 2, . . . , so L0 = 0. Hence, if policymakers can mislead the public, shaping inflation expectations with projections that
are inconsistent with the implemented policy, they can achieve their objectives
(3) to a greater extent than with a credible commitment to the optimal plan.
For comparison, imagine that in any period t = 0 policymakers could mislead the public into believing the policy rule
πt = −β −1 ρ−1 ut
(17)
would be implemented in periods t = 1, 2, 3, . . . , so that the public’s expectations satisfy condition (16), while actually implementing equation (14) in all
periods t = 0, 1, 2, . . . Policymakers would then do better, in terms of the policy objective (3), than with the optimal commitment plan (9)-(10). How is
this possible? While time-inconsistent, at any point in time the optimal commitment plan assumes expectations of future policy are consistent with what
policymakers actually plan to implement, and vice-versa. This restriction is violated in the commitment strategy proposed in the present paragraph. In other
words, policymakers in the standard commitment solution plan to implement
the commitment rule (10) in all future periods, and this rule also shapes expectations of future policy. In the alternative strategy, expectations are assumed
to be shaped by the promised policy (17), but policymakers plan to implement
a different equation (14). Hence, this alternative strategy requires misleading
7
the public period after period, something that might be easier to achieve shaping expectations with projections, rather than through a commitment to a rule.
From a theoretical point of view the public must in both scenarios determine
what policy is implemented, (14) or (17). However, in practice there is a great
difference between determining if a preannounced rule is followed, versus determining whether the implemented policy and (past and present) projections are
mutually consistent. This is especially true when policymakers do not provide
details about how their projections are generated.
If the shock ut and inflation πt were perfectly observable, it would be obvious that projections generated with equation (15) are inconsistent with the
implemented policy (14) whenever ut 6= 0. If the public could observe the implemented policy (14), deduce it by studying past realizations, or derive it by
reproducing the policy problem and policymakers’ logic, inflation expectations
would instead be
Êt πt+1 = Et πt+1 = 0
(18)
for all t = 0, 1, 2, . . . Combined with the implemented policy (14), this implies
lt =
λ 2
u
κ2 t
for t = 0, 1, 2, . . . , and therefore an objective value (3),
∞
Lp0
X
λ
= 2 E0
β t u2t
κ
(19)
t=0
which can be higher or lower than the corresponding loss
Ld0
=
∞
X
E
β t u2t
2 0
2
((1 − βρ) λ + κ )
t=0
λ λ + κ2
(20)
arising with standard discretion (13).6 In particular, Ld0 < Lp0 when
(1 − βρ)2 λ + (1 − 2βρ) κ2 > 0
(21)
while Ld0 ≥ Lp0 otherwise. Hence, even after misleading projections fail to
deceive the public, they can lead to better results than discretion. When expectations become consistent with the implemented policy, so Êt πt+1 = 0, and
policymakers’ misleading projections are no longer believed, it would not be
optimal to implement πt = 0 if changing the implemented policy would have no
impact on expectations (the optimal policy would then be the standard discreThe reduced-form solution under discretion (13) is πt = λ((1 − βρ)λ + κ2 )−1 ut and xt =
−κ((1 − βρ)λ + κ2 )−1 ut , which can be obtained by the method of undetermined coefficients.
6
8
tionary solution (13)). However, policymakers know that if they stop setting
πt = 0, expectations would change, so they have a choice between staying at
the current equilibrium, or moving to the standard discretionary one. When
Lp0 < Ld0 they would be better off remaining in this alternative equilibrium.7
From above (21), we see that Lp0 < Ld0 requires large β and ρ. This is because
the benefit from having Êt πt+1 = 0 is that the effects of a non-zero ut -shock do
not propagate towards the future through expectations, which is more important the more persistent these shocks are, and the more policymakers weight
the future. Lp0 < Ld0 is also more prone to occur the larger is κ and the smaller
is λ, since then a ut -shock has less impact on output and policymakers care
less about output stabilization, respectively (when πt = 0 ∀t the objective (3)
depends only on output stabilization).
In reality inflation is observable, albeit with a lag. However, there are several
factors that can make it difficult for the public to determine whether policymakers’ projections are consistent with the implemented policy. The most obvious
is the uncertainty tied to forecasts, and in particular, the difficulty in accurately predicting macroeconomic variables even just a few quarters ahead. In
addition, there is model uncertainty and misspecification, unknown parameter
values, and the fact that different committee members voting on policy might
favor different models or parameter values. Also, policymakers might not be
able to achieve exactly the inflation rate that they want, contrary to what is
assumed in our model (for example, there might be additional noise that makes
πt differ from zero even when policymakers try their best to make it zero).
Because of these difficulties, policymakers could easily get away with manipulating their projections, and due to the potential gains, they have incentives
to do so. This, however, means that the public should not trust policymakers’
projections, which renders these all but useless.8
4
Optimal time-inconsistent projections
If the public does trust the central bank, what are the optimal projections when
these have to be consistent with the policy implemented, so as to maintain the
public’s confidence? From the perspective of any period t = 0 that is the very
7
This equilibrium is feasible under commitment, and so must be inferior to optimal commitment.
While the public at large may benefit from being mislead in terms of a lower loss from distortions
due to menu costs (3), this ignores the cost of suboptimal decisions individuals might make as a
result of having biased inflation expectations.
8
9
first one in which policymakers project ahead, the optimal initial-period policy
and projections I > 0 periods ahead are given by the sequence {πt }It=0 that
minimizes the loss function (3), subject to the Phillips curve (1). The first-order
conditions are the same as above, equations (6)-(8) for t ≤ I, yielding the same
optimal present action (9) and optimal future actions (10), and projections
(π̂t = πt ), for t = 1, 2, 3, . . . , I. When t = 0 is not the very first period in
which policymakers project ahead, policy and projection equations for periods
t = 0, 1, 2, . . . , I − 1 are given by past optimal projections (10). The optimal
new projection, πI , is the same as all other optimal projections (10).9
As with any commitment to future policy, optimal projections are timeinconsistent. The optimal policy to implement in the contemporary period is
discretion (13), while the optimal policy to promise, or more generally, to shape
policy expectations, is the commitment rule (10).10 Hence, if policymakers
reconsidered in any later period, they would want to implement the optimal
discretionary policy (13), thus deviating from previous projections generated
with the commitment rule (10). With standard commitment, the reason for
deviating is that πτ +1 has an impact on πτ and yτ through expectations Êτ πτ +1
in period τ and prior, but no such impact in period τ + 1 when πτ , yτ and
Êτ πτ +1 have already been realized. Thus, from the perspective of any period
before τ + 1, the commitment rule (10) is optimal for period τ + 1, but from the
perspective of period τ +1, discretion (13) is optimal for period τ +1. Similarly,
in period τ and prior, projections π̂τ +1 affect πτ and yτ through expectations
Êτ πτ +1 , but have no impact on these in period τ + 1. Consequently, in any
period prior to τ + 1, projections π̂τ +1 matter, and thus shape the optimal
policy πτ +1 , when mutually consistent. However, in period τ + 1 the projection
π̂τ +1 has no impact, thus changing the optimal policy for period τ + 1 from
the commitment rule (10) to discretion (13). Hence, optimal projections are
time-inconsistent for the same reason that the optimal plan is.
Because the optimal future policy and projection equations are the same
(10) for all values of I > 0, it follows that according to the present model, there
is no need to project more than one period ahead.11 However, this assumes
9
Studying short-term commitments, Jensen (2013) shows that the Lagrangian approach used
above is valid even when the commitment only applies I periods into the future, whether or not
period t = 0 is the first one in which policymakers commit. Hence, committing, or projecting ahead,
the optimal policy is to follow the standard optimal plan for as long as the commitment lasts.
10
The optimal misleading projections (15) are also time-inconsistent, since policymakers would
rather implement (14). However, contrary to the projections studied in the present section, misleading projections are not even consistent from a static point of view.
11
This is analogous to the finding that there is nothing to be gained from continuously committing
10
that the public can replicate the central bank’s policy problem, and thus figure
out that future policy, and hence projections and expectations, will satisfy the
commitment rule (10) also beyond the initial I periods. If instead the public
expects discretion (13) after the first I periods, policymakers might be better
off the more periods I into the future they project. Hence, how far into the
future it is advantageous to project depends on one’s assumptions about the
public’s expectations beyond policymakers’ projections.
5
Optimal time-consistent projections
When projections are not time-inconsistent, or used to mislead the public, there
is nothing to be gained by policymakers from providing projections, unless the
public’s expectations are inconsistent with the implemented policy, or underlying economics. In the ideal case where the public can replicate the policy
problem perfectly, and thereby form expectations, they would correctly anticipate the optimal discretionary policy (13). Consequently, their own forecasts
of macroeconomic aggregates would be equivalent to policymakers’ projections.
A case in which projections can play a role, even when these are timeconsistent and non-misleading, is when the public’s expectations are rational,
but more noisy than policymakers’. As an example, imagine
λ
Êt πt+1 = − Et xt+1 + γvt
κ
(22)
vt = φvt−1 + et
(23)
where
and et is white noise, while the implemented policy is optimal discretion (13).
With such expectations, the value of the policy objective (3) is
Ln0
=
Ld0
+
γ 2 λ λ + κ2
((1 − βφ) λ + κ2 )
2 E0
∞
X
β t vt2
(24)
t=0
which is larger than Ld0 , more so the further γ is from zero. Hence, if policymakers’ projections contribute toward reducing the degree of noise γ in the
public’s expectations, the policy objective (3) would improve.
When the public’s expectations are inconsistent with the implemented polmore than one period ahead (Jensen (2013)).
11
icy, the issue is more complicated. Assuming, as an example, that expectations
Êt πt+1 = ηEt xt+1
(25)
where η is a constant, the value of the policy objective (3) would be
Li0
=
λ λ + κ2
(λ + βρκη + κ2 )
2 E0
∞
X
β t u2t
(26)
t=0
when discretion (13) is implemented. It follows that when η is between −λ/κ
and (βρλ − 2λ − 2κ2 )/(βρκ), Ld0 < Li0 , and the policy objective (3) would be
lower if the public’s expectations were corrected so as to be consistent with
the implemented policy (η = −λ/κ). For all other values of η, Ld0 > Li0 , and
outcomes (3) are better when the public’s expectations remain inconsistent. Expectations can be inconsistent with implemented policy (13) in different ways
other than explored above (25). While arbitrary, the case above is simple, yet
general enough to illustrate that the result is ambiguous. Correcting the public’s inconsistent expectations may improve or deteriorate economic outcomes.
6
FOMC projections
Since its 2007 October meeting, the FOMC has published projections of real
GDP growth, unemployment rate and inflation rate after many of its policy
meetings. These were released with its policy statement, at the Chairman’s
press conference within hours of the meeting, or in the minutes released three
weeks afterwards.12 Projections have been produced about four times a year,
for the January, March/April, June and October/November/December FOMC
meetings, with the schedule changing slightly over the years.13 The projections
cover the concurrent year, and one and two years into the future. Each FOMC
member, whether or not voting in that particular meeting, submits a projection,
yielding 16-19 individual projections for each meeting. The detailed data is not
12
From 1979 to 2007, projections were generated twice a year (January and June in preparation
for Congressional testimony in February and July). Starting April 2011, projections were released
in conjunction with the Chairman’s post-meeting press conference, and are more recently also summarized in the policy statement published immediately after the meeting. In 2012, the FOMC also
started providing projections of the target federal funds rate, but none of the surveys forecast it.
13
From 2008 to 2011, projections were provided in January, April, June and October/November.
In 2012 they were provided in January, April, June, September and December. From 2013 to 2015
they were provided in March, June, September and December.
12
publicly available until transcripts are released five years later. Instead, the
FOMC makes immediately available the range of the individual projections, and
what it calls the central tendency numbers, that is, the minimum and maximum
projected values, after having eliminated the three lowest and three highest
ones. For comparison with other forecasts, we plot also the midpoint of these
central tendency numbers, which is the closest we can get, with the immediately
available data, to the median forecast from the comparison surveys.14
The FOMC projects the average civilian unemployment rate in the fourth
quarter of each year, and fourth-quarter to fourth-quarter growth rates of real
GDP and prices for personal consumption expenditures (PCE). This makes it
difficult to compare with survey forecasts, which typically focus on annual averages. The FOMC’s particular, and changing, schedule further complicates comparison, as the surveys follow more regular patterns. For example, the survey
of professional forecasters (SPF), our main comparison survey, operates with
forecast deadlines in the second week of February, May, August and November.
Consequently, the FOMC’s January and March projections are both compared
with the SPF February forecast, the FOMC’s April and June projections are
compared with SPF May, the FOMC’s September value is compared with SPF
August, while the FOMC’s October, November and December projections are
all compared with the SPF November forecast. This way the predictions are at
most a month apart.15 The dates used in the figures below are the second day
of the FOMC meetings for which the corresponding projections were produced.
There is no information available on the exact dates the projections and forecasts were actually generated. Initial FOMC projections are due by the end of
the day on the Friday before the FOMC meeting, but may be revised at any
time until the beginning of the session on the second day of the meeting (previously the final deadline was the day after the meeting ended). Comparing the
FOMC’s projections with other predictions is crucial, since other forecasters
would not have incentives to sway the public’s expectations in the same way
that policymakers might have.16
14
If projections affect expectations, it should be through those released immediately. From
September 2015, the releases include median projections, so we use these instead of midpoints.
15
Even if predictions are close in time, they can vary significantly if new data is released in-between.
Estimates of GDP and personal expenditures are released at the end of the month. Unemployment
data is released at the beginning of each month. Not all survey respondents exhaust the deadline,
and some might not use the most up-to-date information available. Comparing with Greenbook
projections, Romer and Romer (2000) find that these timing differences do not appear to have any
impact.
16
Both the SPF and the Livingston surveys include employees at the Federal Reserve and its
13
Figure 1 plots projections and forecasts of annual real GDP growth (fourth
quarter to fourth quarter) for 2008 to 2015, together with the realized values
(as reported by the FOMC). Each panel plots forecasts for a given year, against
the time at which these were produced, together with the realized value. The
FOMC’s long-term growth projections tend to start out too optimistic, and are
gradually adjusted down. This initial overshooting may partly be due to the
low rates of growth experienced in the years studied, though growth was about
average in 2010, 2013 and 2014. Unfortunately, there are not many comparable
forecasts. The SPF forecasts average annual growth three years ahead, but
only a year ahead on a quarterly basis. In addition, the survey actually asks
for GDP levels, so the implied fourth-quarter to fourth-quarter growth rate
can only be computed with certainty for the first and fourth quarter surveys.
Similarly, the Livingston survey asks participants to forecast GDP only six and
twelve months ahead, yielding only one comparable value per year.
FOMC projections of unemployment for 2008-12, which can be seen in figure
2, also started out too optimistic. This may be due to the exceptionally high
and persistent unemployment in the years studied, which made the three-year
ahead forecasts rise from 5% to 7.25% between 2010 and 2012. For 2013-15,
FOMC forecasts start out too pessimistic. FOMC projections of unemployment
do not appear to differ systematically from the SPF forecasts, except maybe
for 2013 and 2014, for which they are always lower. However, for these two
years the SPF forecasts did worse than the FOMC’s.
Figure 3, which portrays projections and forecasts of PCE inflation, shows
that the FOMC appears to systematically project lower inflation than the SPF
since 2011, and not always correctly so (2011 and 2012). The same applies
also for the long-run forecasts for 2010. To test this statistically, note that if
the FOMC and survey expectations were drawn from the same distribution, so
that they were both noisy observations of the same underlying expectations,
we would expect one to be higher than the other half of the time, and viceversa, whenever they differ. This would make the (two-sided) probability of one
being above the other for exactly x consecutive observations equal 2 × ( 12 )x =
21−x , which makes the probability of x or more such consecutive observations
21−x +21−(x+1) +21−(x+2) +21−(x+3) +. . . = 21−x (1+2−1 +2−2 +2−3 +. . . ) = 22−x .
Hence, seven or more consecutive observations of one being above the other
would have less than .05 probability of occurring, making us reject that they
member banks, and thus possibly some of the people responsible for FOMC projections.
14
are draws from the same distribution with the usual 95% confidence. Ten or
more such observations, as we observe for 2013, 2014 and 2015, would each occur
with less than .004 probability.17 The probability of 23 consecutive observations
of one below the other, as we see for 2013 and 2014 jointly is 1.2 × 10−7 .
A formal Binomial test reveals that FOMC projections being lower than the
corresponding SPF forecasts in 74 out of 87 cases, as we observe for inflation
from 2008 to 2015, only occurs with a probability 1.6 × 10−11 , if the underlying
distributions were truly the same.18
On average over the period 2008-15, the FOMC midpoint projection for
inflation was .20 percentage points below the SPF median forecast. In terms of
absolute deviations, so that positive and negative deviations do not cancel each
other out, the average deviation was .27 percentage points. Annual averages
of absolute deviations are all within the .22-.30 range, except for .42 in 2008,
and ignoring this outlier, show no evidence of declining over time. The single
largest deviations are of a whole percentage point in 2008, and .65 percentage
points in both 2012 and 2013.19 As figure 3 shows, there is much more variation
in the magnitude of deviations between forecasts and projections than in their
sign.
While the comparison data is limited, FOMC projections often do better
than those of the SPF, though not always. Hence, FOMC projections do not
appear to be misleading. Why then, do the two differ systematically? In particular, why are survey respondents’ inflation forecasts systematically higher
that those of the FOMC? In FOMC discussions (see transcripts from January
and June 2007), several members voice qualms about long-run projections of
inflation that are too far away from the unofficial 1-2.5% target rate, particularly since the FOMC descriptions state that “Longer-run projections represent
each participant’s assessment of the rate to which each variable would be expected to converge under appropriate monetary policy and in the absence of
further shocks to the economy.” From this point of view, the FOMC’s inflation
projections should convey its commitment to low inflation, or at least not be
inconsistent with such a commitment, while survey forecasts do not face this
constraint. However, for the years studied, both FOMC and SPF long-run fore17
The same applies to the unemployment forecasts for 2013 and 2014 jointly.
The FOMC’s unemployment projections were below those of the SPF in 26 out of 41 occasions,
while for GDP growth it was above those of the SPF’s 9 out of 15 times, which are not statistically
significant differences according to the Binomial test.
19
Deviations with the Livingston survey median inflation are consistently greater or equal to the
deviations with the SPF for the few observations that match up.
18
15
casts are mostly within the target range, making such a constraint irrelevant.20
Whatever the reason for the systematic deviations, it is clear that, despite its
good track record, the FOMC’s projections have failed to dictate the public’s
inflation expectations, or have at least not allowed it to control these perfectly.
The FOMC has motivated its move towards providing projections by arguing that this leads to increased transparency. This is reflected in transcripts
from FOMC meetings in 2007 (January), where the idea of publishing projections, and different options for doing so, were discussed at length. Some other
motivations were that it would increase the Fed’s credibility and accountability,
strengthen its commitment to price stability, and that it could make monetary
policy more effective. Arguably, the projections give the public an idea about
where FOMC members believe the economy is heading, thus providing a rationale for their policy decisions. In this sense, it may be useful to make immediately available to the public also the projections generated by the staff at the
Board of Governors (currently released after five years), which apparently carry
considerable weight in the policy discussion, and all other information available
to the committee.21 It would also be useful to specify how the projections are
generated. On the other hand, Amato et al. (2002), Geraats (2002) and Jensen
(2002) argue that transparency can sometimes deteriorate economic outcomes.
Moreover, FOMC’s projections give insight into any divisiveness that might
exist within the committee, which can increase uncertainty about its future
actions, especially since the FOMC states that “Each participant’s projections
are based on his or her assessment of appropriate monetary policy.”
Meeting transcripts (January and June 2007) show that the FOMC has
discussed the influence its projections can have on the public’s expectations.
In particular, some members expressed concerns about people taking the projections as a commitment by the FOMC to realize the projected values. As
mentioned above, there also seems to be some concern that the public’s beliefs
about the Fed’s willingness to defend its goal of price stability, relative to that
of stimulating economic activity, might be affected by its projections, especially
those of long-run inflation.
20
Moreover, in 2008, the only year in our sample inflation expectations rose above 2.5%, FOMC
projections were actually higher than the SPF forecasts. Inflation expectations dropped below 1%
in both 2009 and 2015. In the first year, FOMC projections were higher than those of the SPF, but
the situation was the opposite in 2015.
21
Romer and Romer (2000) argue that the Greenbook forecasts produced by the Fed staff were
more accurate than private-sector forecasts for the years they study (1965-1991).
16
7
Conclusions
It is unclear whether policymakers’ projections, and forward-guidance in general, have much effect on the public’s expectations. Our theoretical analysis
shows that policymakers have incentives to use projections to mislead the public, so that these should be viewed with skepticism. Consistent with this, we
find that survey expectations differ significantly from projections, thus rejecting the idea that these allow for controlling the public’s expectations, at least
in the FOMC’s case over the 2008-15 period. However, we cannot conclude
that projections have no influence on expectations. At the same time, there
is a lack of studies on the impact of projections on economic outcomes. One
might think that policymakers’ have nothing to lose, and that in the worst case,
projections are just ignored. However, we find that correcting the public’s inconsistent expectations can sometimes deteriorate outcomes. Furthermore, it
is easy to imagine that poor projections, especially if worse than those of other
forecasters, could deteriorate the public’s trust in policymakers’ abilities to understand and manage the economy, which could bring about expectations that
lead to worse economic outcomes. Less favorable expectations can arise even
when projections are accurate, by revealing policymakers’ true preferences, or
by making the public draw incorrect inferences about these. For example,
policymakers’ projections could reveal, or make the public believe, that policymakers are not as hawkish on inflation as previously thought, leading to higher
inflation expectations and worse economic outcomes.
8
References
Amato, J. D., Morris, S. and Shin, H. S. (2002) “Communication and Mon-
etary Policy.” Oxford Review of Economic Policy 18: 495-503.
Assenmacher-Wesche, K. (2006) “Estimating Central Banks preferences from
a time-varying empirical reaction function.” European Economic Review 50:
1951-1974.
Barro, R. J. and Gordon, D. B. (1983a) “A Positive Theory of Monetary
Policy in a Natural Rate Model.” Journal of Political Economy 91: 589-610.
Barro, R. J. and Gordon, D. B. (1983b) “Rules, discretion and reputation
in a model of monetary policy.” Journal of Monetary Economics 12: 101-121.
Blinder, A. S., Ehrmann, M., Fratzscher, M., De Haan, J., and Jansen, D. J.
(2008) “Central bank communication and monetary policy: A survey of theory
17
and evidence. Journal of Economic Literature 46: 910-945.
Bowen, A. (1995) “British Experience with Inflation Targetry” in L. Leiderman and L. E. O. Svensson, eds. Inflation Targets, CEPR, London.
Calvo, G. A. (1983) “Staggered Prices in a Utility Maximizing Framework.”
Journal of Monetary Economics 12: 383-398.
Clarida, R., Gali, J. and Gertler, M. (1999) “The science of monetary policy:
A new Keynesian perspective.” Journal of Economic Literature 37: 1661-1707.
Cukierman, A. and Meltzer A. H. (1986) “A Theory of Ambiguity, Credibility, and Inflation under Discretion and Asymmetric Information.” Econometrica 54: 10991128.
Currie, D. and Levine, P. (1993) Rules, Reputation and Macroeconomic
Policy Coordination, Cambridge University Press.
Dennis, R. (2004) “Inferring Policy Objectives from Economic Outcomes.”
Oxford Bulletin of Economics and Statistics 66: 735-764.
Geraats, P. (2002) “Central Bank Transparency.” The Economic Journal
112: 532-565.
Givens, G. E. (2012) “Estimating Central Bank Preferences under Commitment and Discretion.” Journal of Money, Credit and Banking 44: 1033-1061.
Hall, R. E. (1985) “Monetary Strategy with an Elastic Price Standard.”
in Price Stability and Public Policy: A Symposium Sponsored by the Federal
Reserve Bank of Kansas City, Federal Reserve Bank of Kansas City.
Hall, R. E. and Mankiw, N. G. (1994) “Nominal Income Targeting” in N.
G. Mankiw, ed. Monetary Policy, University of Chicago Press, Chicago.
Jensen, C. (2013) “The Gains from Short-term Commitments,” Journal of
Macroeconomics 35: 14-23.
Jensen, H. (2002) “Optimal Degrees of Transparency in Monetary Policymaking.” Scandinavian Journal of Economics 104: 399-422.
King, M. A. (1994) “Monetary Policy in the UK.” Fiscal Studies 15:109-128.
Kydland, F. E. and Prescott, E. C. (1977) “Rules Rather than Discretion:
The Inconsistency of Optimal Plans.” Journal of Political Economy 85: 473492.
J. J. Rotemberg, eds., NBER macroeconomics annual 1994. Cambridge,
MA. MIT Press, 1994, pp. 1357.
Romer, C. D. and Romer, D. H. (2000), “Federal Reserve Private Information and the Behavior of Interest Rates.” American Economic Review 90:
429-457.
18
Rotemberg, J. J. (1982) “Monopolistic Price Adjustment and Aggregate
Output.” Review of Economic Studies 49: 517-531.
Schorfheide, F. (2008) “DSGE Model-Based Estimation of the New Keynesian Phillips Curve.” Federal Reserve Bank of Richmond Economic Quarterly
94: 397-433.
Söderström, U., Söderlind, P. and Vredin, A. (2005) “New-Keynesian Models and Monetary Policy: A Re-examination of the Stylized Facts.” The Scandinavian Journal of Economics 107: 521-546.
Svensson, L. E. O. (1997) “Inflation Forecast Targeting: Implementing and
Monitoring Ination Targets.” European Economic Review 41: 1111-1146.
Svensson, L. E. O. (1999) “Inflation targeting as a monetary policy rule.”
Journal of Monetary Economics 43: 607-654.
Svensson, L. E. O. (2005) “Monetary Policy with Judgment: Forecast Targeting.” International Journal of Central Banking 1: 1-54.
Svensson, L. E. O. and Woodford, M. (2002) “Indicator Variables for Optimal Policy.” Journal of Monetary Economics, 50, 691-720.
Woodford, M. (1999) “Commentary: How Should Monetary Policy Be Conducted in an Era of Price Stability,” in New Challenges for Monetary Policy,
Federal Reserve Bank of Kansas City: 277-316.
Woodford, M. (2003) Interest and Prices: Foundations of a Theory of Monetary Policy. Princeton University Press.
Woodford, M. (2004), “Inflation targeting and optimal monetary policy.”
Review, Federal Reserve Bank of St. Louis July: 15-42.
Woodford, M. (2007) “The Case for Forecast Targeting as a Monetary Policy
Strategy.” Journal of Economic Perspectives 21: 3-24.
19
Fed min/max projections
Fed midpoint/median projection
Realized value
SPF forecast
Livingston forecast
g
Figure legend
20
2.0%
2008
1.0%
2.5%
1.5%
0.0%
0.5%
‐1.0%
‐0.5%
‐2.0%
‐1.5%
‐1 5%
‐3.0%
‐2.5%
3.5%
2010
3.0%
2.5%
2.0%
1.5%
5.0%
4.5%
4.0%
3.5%
3.0%
2 %
2.5%
2.0%
1.5%
1.0%
2012
2014
3.0%
5.0%
4 5%
4.5%
4.0%
3.5%
3.0%
2.5%
2.0%
1.5%
5.0%
4.5%
4.0%
%
3.5%
3.0%
3 0%
2.5%
2.0%
1.5%
2011
2013
4.0%
4.0%
3.5%
2009
3.5%
2015
3.0%
2.5%
2.5%
2.0%
2 0%
2.0%
1 5%
1.5%
Figure 1: Projections and forecasts of real GDP growth.
21
7.5%
7.0%
2008
6.5%
6.0%
5.5%
5.0%
11.0%
10.0%
9.0%
8.0%
7.0%
6.0%
5.0%
4.0%
9.0%
9 0%
2011
8.0%
7.0%
6.0%
5.0%
8.5%
2012
2013
8.0%
8.0%
7.5%
7.5%
7 5%
7.0%
7.0%
6.5%
6.5%
8.0%
7.5%
2009
10.0%
2010
9.0%
8.5%
11.5%
10.5%
9.5%
8.5%
7.5%
6.5%
5.5%
5 5%
4.5%
7.0%
2014
6.5%
7.0%
6.0%
6.5%
5.5%
6.0%
5.0%
5 5%
5.5%
4 5%
4.5%
2015
Figure 2: Projections and forecasts of unemployment.
22
4.5%
4.0%
3.5%
3.0%
2.5%
2.0%
1 5%
1.5%
1.0%
2.5%
2008
2.0%
2009
1.5%
1.0%
0.5%
0.0%
2.5%
3.5%
2010
2.0%
3.0%
2011
2.5%
2.0%
1.5%
1.5%
1.0%
1.0%
0.5%
2.5%
2.5%
2012
2013
2.0%
2.0%
1.5%
1.5%
1.0%
1.0%
0.5%
2.5%
2.5%
2014
2.0%
1.5%
2.0%
2015
1.5%
1.0%
0.5%
1 0%
1.0%
0 0%
0.0%
Figure 3: Projections and forecasts of PCE inflation.
23