The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Federal Reserve Bank of Chicago
Macroeconomic Effects of
Federal Reserve Forward Guidance
Jeffrey R. Campbell, Charles L. Evans,
Jonas D.M. Fisher, and Alejandro Justiniano
WP 2012-03
jeffrey r. campbell
charles l. evans
Federal Reserve Bank of Chicago
Federal Reserve Bank of Chicago
jonas d. m. fisher
alejandro justiniano
Federal Reserve Bank of Chicago
Federal Reserve Bank of Chicago
Macroeconomic Effects
of Federal Reserve Forward Guidance
ABSTRACT A large output gap accompanied by stable inflation close to its
target calls for further monetary accommodation, but the zero lower bound
on interest rates has robbed the Federal Open Market Committee (FOMC) of
the usual tool for its provision. We examine how public statements of FOMC
intentions—forward guidance—can substitute for lower rates at the zero bound.
We distinguish between Odyssean forward guidance, which publicly commits
the FOMC to a future action, and Delphic forward guidance, which merely
forecasts macroeconomic performance and likely monetary policy actions.
Others have shown how forward guidance that commits the central bank to
keeping rates at zero for longer than conditions would otherwise warrant can
provide monetary easing, if the public trusts it. We empirically characterize
the responses of asset prices and private macroeconomic forecasts to FOMC
forward guidance, both before and since the recent financial crisis. Our results
show that the FOMC has extensive experience successfully telegraphing its
intended adjustments to evolving conditions, so communication difficulties do
not present an insurmountable barrier to Odyssean forward guidance. Using
an estimated dynamic stochastic general equilibrium model, we investigate
how pairing such guidance with bright-line rules for launching rate increases
can mitigate risks to the Federal Reserve’s price stability mandate.
F
rom the onset of the financial crisis and through the Great Recession
and ensuing modest recovery, the Federal Open Market Committee
(FOMC) of the Federal Reserve has commented upon the likely duration
of monetary policy accommodation in the formal statement that follows
each of its meetings. In December 2008 it said, “The Committee anticipates
that weak economic conditions are likely to warrant exceptionally low
Brookings Papers on Economic Activity, Spring 2012
Copyright 2012, The Brookings Institution
1
2
Brookings Papers on Economic Activity, Spring 2012
levels of the federal funds rate for some time.” In March 2009, when the
first round of large-scale purchases of Treasury securities was announced,
“an extended period” replaced “some time” in the formal statement. The
August 2011 FOMC statement gave specificity to “an extended period” by
announcing that the committee expected the funds rate to remain exceptionally low until “at least . . . mid-2013.” The January 2012 statement
lengthened the anticipated period of exceptionally low rates even further,
to “late 2014,” language that remained in the March 2012 statement. Such
communications of monetary authorities’ intentions are referred to as forward guidance.
The nature of this most recent forward guidance by the FOMC is the
subject of substantial debate. Studies by Paul Krugman (1999) and by Gauti
Eggertsson and Michael Woodford (2003) before the recent episode and by
Iván Werning (2012) more recently suggest that a monetary policymaker
encountering the zero lower bound (ZLB) on the policy interest rate can
stimulate current aggregate demand by credibly promising to keep the rate
at zero longer than required by economic conditions and thereby creating
an economic boom in the future. One might interpret “late 2014” as such a
credible promise, but one also might interpret it as merely describing what
the FOMC’s policy reaction function would prescribe if current forecasts
of sluggish economic activity and low inflation through that date come to
pass.1 “Late 2014” predicts unusually accommodative policy whenever the
underlying policy reaction function would dictate an earlier “liftoff ” of the
funds rate from zero given the identical conditioning data.
Motivated by these competing interpretations of “late 2014,” we distinguish between two kinds of forward guidance. Delphic forward guidance publicly states a forecast of macroeconomic performance and likely
or intended monetary policy actions based on the policymaker’s potentially superior information about future macroeconomic fundamentals
and its own policy goals.2 Such forward guidance presumably improves
macroeconomic outcomes by reducing private decisionmakers’ uncertainty.
1. Since one of the authors regularly attends meetings of the FOMC, it may be tempting
just to ask him this question directly. The vantage point of this paper is a research inquiry:
how can these questions be answered from the standpoint of economic researchers with only
publicly available information?
2. The classical Delphic oracle famously made ambiguous utterances. We do not mean
“Delphic” in this sense. We use the term simply to describe FOMC statements about the
future.
campbell, evans, fisher, and justiniano
3
Importantly, however, it does not publicly commit the policymaker to a
particular course of action. Odyssean forward guidance, in contrast, does
publicly commit the policymaker, just as Odysseus committed himself to
staying on his ship by having himself bound to the mast. Tying one’s hands
in the face of an uncertain future might seem like a foolish sacrifice for no
apparent gain, but economic fluctuations routinely present opportunities
for monetary policy to benefit from issuing Odyssean forward guidance.
The reason is that by so doing, policymakers can change public expectations of their actions tomorrow in a way that improves macroeconomic
performance today.3
Nevertheless, the implementation of Odyssean policy faces a fundamental
challenge. When the appointed time for action arrives, any beneficial effects
of the policy’s anticipation will be bygones that nothing can change. Therefore, both the monetary policymaker and the public will at that later time
prefer a policy that addresses only the present circumstances and ignores
the beneficial effects of its anticipation on past macroeconomic performance. For example, when it comes time to keep an earlier promise to raise
aggregate demand, the FOMC will be concerned about its price stability
mandate and, acting as it has always done in normal times, will not want to
follow through.4 Just as Odysseus anticipated that on hearing the Sirens’ song
he would regret his commitment to stay aboard his ship, so might monetary
policymakers anticipate regretting their commitment to ease policy. If the
public understands this and therefore believes that such promises will not be
kept, they will not have the intended effect. Odysseus could use the rope that
bound him to the mast to enforce his commitment. Lacking such an enforcement mechanism, monetary policymakers must rely on their reputations for
accuracy and honesty to make their commitments credible.
The Odyssean monetary policies elucidated by Krugman, Eggertsson and
Woodford, and Werning have inspired several recent proposals to provide
more accommodation at the ZLB. The more aggressive policy alternatives
that have been proposed include Evans’s (2012) state-contingent price-level
targeting, nominal income targeting as advocated by Christina Romer,5 and
3. Romer and Romer (2000) and Ellingsen and Söderström (2001) characterize forward
guidance similarly.
4. This is an example of a time-inconsistent policy, first considered by Kydland and
Prescott (1977).
5. Christina D. Romer, “Dear Ben: It’s Time for Your Volcker Moment,” The New York
Times, October 29, 2011.
4
Brookings Papers on Economic Activity, Spring 2012
conditional economic thresholds for exiting the ZLB as proposed by Evans
(2011). The main challenge facing the FOMC in implementing any of these
policies is convincing the public that it will follow through on the promised
future course of action. This paper sheds light on the FOMC’s ability to
meet this challenge and on the possible benefits of doing so.
The FOMC has used forward guidance implicitly, through speeches and
testimony by its members, and explicitly, through formal committee statements, since long before the financial crisis, so the question of whether
the FOMC can clearly communicate its future policy intentions can be
addressed empirically. Accordingly, the first part of this paper examines
data from before and after the crisis, to measure the impact that FOMC
communications have had on private expectations. We begin by studying
market responses to FOMC statements, building on prior work by Refet
Gürkaynak, Brian Sack, and Eric Swanson (2005). Those authors follow
Kenneth Kuttner (2001) by analyzing changes in prices on federal funds
rate futures in short windows of time surrounding the release of FOMC
statements. Using a sample from June 1991 through December 2004,
Gürkaynak and his coauthors find that FOMC statements are associated
with significant effects, both on federal funds futures prices and on Treasury yields, that are not due to surprise changes in the federal funds target itself. That is, their results show that market participants believe that
FOMC statements contain reliable information about future monetary policy actions. We verify that these findings continue to hold when the sample
is extended to July 2007, just before the crisis.
One might doubt the relevance of these findings for the present situation, because the attainment of the ZLB has robbed the FOMC of its
principal policy lever. But evidence exists that the FOMC can still exert
influence in the presence of a binding ZLB. Focusing on FOMC communications about its recent large-scale asset purchases, known as QE1
and QE2, Joseph Gagnon and coauthors (2010) and Arvind Krishnamurthy and Annette Vissing-Jorgensen (2011) provide evidence of significant asset price effects since the crisis. To complement these studies
and provide more assurance that forward guidance unaccompanied by
material policy action can move asset prices, we apply Gürkaynak and
his coauthors’ methodology to FOMC statements since the crisis and find
results similar to theirs.
FOMC actions that influence asset prices are merely means toward the
end of fulfilling the Federal Reserve’s dual mandate of maximum sustainable employment and price stability. To evaluate the contributions of
campbell, evans, fisher, and justiniano
5
FOMC statements toward this ultimate goal, we examine how revisions to
the Blue Chip consensus forecasts of the unemployment rate and consumer
price index (CPI) inflation respond to the policy innovations identified by
Gürkaynak and others (2005). For the sample period February 1994 to June
2007, a positive innovation to future federal funds rates is associated with
decreases in unemployment forecasts for the subsequent 3 quarters and
with higher forecasts of CPI inflation in the current and subsequent quarters. We never find a statistically significant reaction of either forecast that
is of the “correct” sign, that is, one that indicates a New Keynesian response
to an exogenous policy shock. From this we conclude that the monetary
policy surprises identified with high-frequency data have a substantial Delphic component, despite the fact that the methodology of Gürkaynak and
others inherently controls for publicly known macroeconomic fundamentals. That is, professional forecasters infer that the FOMC’s unexpected
policy adjustments are responses to nonpublic information that the FOMC
possesses regarding the future strength of the economy.6 We find qualitatively similar results for the crisis period, but the estimates are too imprecise to allow firm quantitative conclusions.
The FOMC does not rely solely on postmeeting public statements to
communicate its policies. To get a broader perspective on the influence of
FOMC communications on private expectations, we proceed to examine
monetary policy surprises identified from a simple interest rate rule like
those of John Taylor (1993, 1999) and David Reifschneider and John Williams (2000). Using the Blue Chip forecasts and interest rate futures prices
aggregated to the quarterly level, we estimate such a rule and decompose
its residual into the part revealed when the spot policy rate is set and the
parts revealed to the public in the prior 4 quarters.
We highlight here four results based on data from 1996 through 2007.
First, the standard deviation of the expected interest rate 4 quarters out
minus its value from the rule is only 9 basis points (bp). Thus, the rule
describes medium-run forecasts of FOMC behavior extremely well. Apparently, the FOMC has been successful in communicating its typical behavior to the public. Although this need not reflect an Odyssean commitment,
it is observationally equivalent to one. Second, the FOMC telegraphs
6. Such information might reflect the Federal Reserve staff’s possibly superior ability to
process incoming data. It does not have to involve proprietary access to data or information
held only by the FOMC about its future policy intentions.
6
Brookings Papers on Economic Activity, Spring 2012
40 percent of its deviations from the interest rate rule exactly 1 quarter
in advance and another 40 percent 2 or more quarters in advance. Third,
the identified forward guidance residuals have much stronger effects on
asset prices than do surprises of the type described by Gürkaynak and
others (2005). For example, a 1-bp innovation to next quarter’s expected
federal funds rate moves both the 2-year and the 5-year Treasury rate by
about 2 bp. The corresponding effects estimated with the methodology
of Gürkaynak and others are under 1 bp. Fourth, the identified forward
guidance residuals are negatively correlated with unemployment forecast
revisions and positively correlated with inflation forecast revisions, just
like the statement date–based shocks in Gürkaynak and others (2005).
Apparently, the residuals reflect, at least in part, anticipated deviations
from the policy rule that nevertheless are motivated by recent news of
economic fundamentals. Phrased differently, the FOMC’s behavior has
been history dependent: the committee reacts more aggressively to economic weakness revealed only shortly before its onset than to weakness
foreseen 4 quarters in advance.
The estimated effects of FOMC forward guidance on asset prices and
private forecasts suggest that the FOMC has had some success in communicating its future intentions to the public.7 This suggests that communication difficulties do not present an insurmountable barrier to monetary
policy based on Odyssean forward guidance. The second part of our paper
investigates the consequences of interpreting the “late 2014” statement
language as Odyssean forward guidance that implements the policy recommendations of Eggertsson and Woodford (2003) and others. There are
legitimate concerns that forward guidance of this kind places the FOMC’s
mandated price stability goal at risk. We consider these concerns by forecasting the path of the economy with the present forward guidance and
subjecting that forecast to two upside risks: higher inflation expectations
and faster deleveraging by households and firms.
This policy analysis uses a medium-scale dynamic stochastic general
equilibrium (DSGE) model adapted from Justiniano, Giorgio Primiceri,
and Andrea Tambalotti (2011) at the Federal Reserve Bank of Chicago.
The model strongly resembles other medium-scale DSGE models in the
7. Both our inferences of forward guidance and those from the more familiar event-study
approach use market prices to measure the quantitative content of FOMC communication. In
standard models the process of communication is transparent and frictionless, so it is tempting to suppose that the FOMC can fine-tune its statements to achieve any desired market
impact. However, one must acknowledge frictions in the communication process that make
market responses to FOMC statements unpredictable to the FOMC itself.
campbell, evans, fisher, and justiniano
7
literature and is very similar to models used at central banks around the
world.8 Importantly for our purposes, it embodies the basic mechanisms
that make forward guidance attractive at the ZLB.
Evans (2011) has proposed that the FOMC pledge to begin lifting its
policy rate from zero if either the unemployment rate falls below 7 percent
or expected inflation over the medium term rises above 3 percent. This “7/3”
threshold rule is designed to maintain low interest rates even as the economy
begins expanding on its own (as prescribed by Eggertsson and Woodford
2003), while providing safeguards against unexpected developments that
may put the FOMC’s price stability mandate in jeopardy. Our policy analysis suggests that such conditioning, if credible, could be helpful in limiting
the inflationary consequences of a surge in aggregate demand arising from
an early end to the deleveraging observed since the financial crisis.
I. FOMC Statements and Private Expectations
The FOMC’s use of forward guidance since long before the financial
crisis makes it possible to assess empirically its ability to communicate its future policy intentions. In this section we do so by applying
the methodology of Gürkaynak, Sack, and Swanson (2005; GSS henceforth). They use high-frequency data on prices of federal funds futures
and Eurodollar futures contracts to measure unanticipated changes in
expected future spot interest rates associated with FOMC statements.
Two estimated factors, a target factor that moves the current policy rate
and a path factor that moves only expected future rates, account for most
of these changes. GSS show that yields on longer-duration Treasury
notes respond substantially to the path factor.
We extend the GSS analysis in three ways. First, we examine the
responses of yields on corporate bonds to the factors and confirm that a positive realization of the path factor raises not only expected future policy rates
but corporate borrowing rates as well. That is, forward guidance influences
interest rates that are directly relevant for private investment decisions. Second, we examine how revisions to professional forecasts of unemployment
and CPI inflation respond to the factors. If the public and the FOMC were
equally well informed about macroeconomic fundamentals, then the factors
must reflect the revelation of FOMC policy preferences. In that case one
would expect forecast revisions to match the equilibrium response to an
8. The FOMC’s minutes for the June 2011 meeting describe a discussion of DSGE
models within the Federal Reserve System at that meeting.
8
Brookings Papers on Economic Activity, Spring 2012
unanticipated monetary policy shock. Instead, however, we find that the statistically significant responses all have the sign opposite to that predicted by
the standard New Keynesian model: unanticipated increases in the path factor lead to decreases in expected unemployment and increases in expected
inflation. From this we conclude that professional forecasters believe that
FOMC policy surprises contain useful and otherwise unavailable macroeconomic information—that is, they have a Delphic component. Third, we
extend the sample period so as to examine FOMC announcements since
the onset of the financial crisis in August 2007. Here the relatively small
sample makes our estimates of professional forecasters’ responses to surprise monetary policy moves too imprecise to allow firm conclusions, but
the estimates of asset price responses remain accurate enough to show that
they differ little from their precrisis values.
I.A. Forward Guidance before the Financial Crisis
Glenn Rudebusch and Williams (2008) describe the modern history of
explicit forward guidance before the financial crisis. From 1983 to 1999 the
FOMC’s views about the future policy path were put to a vote at each meeting. The vote was on the expected direction of future changes in the stance
of policy between meetings. However, this information was made public
only after the following meeting, when it was outdated and presumably of
limited use to the public. In February 1994 the FOMC began issuing immediately after each meeting a statement describing the current policy stance,
and in May 1999 it began including explicit language about the future
stance of policy in these statements. The first of these forward-looking
statements read in part as follows: “The Committee . . . adopted a directive
that is tilted toward the possibility of a firming in the stance of monetary
policy.” The language intended to guide expectations has changed over
time as the FOMC has sought ways of maintaining transparency without
confusing markets, and as it has adjusted to the evolving policy environment. But language of one form or another describing the expected future
stance of policy has come to be a fixture of these statements.9
9. Here are some examples. At the start of 2000, the direct signals of policy inclinations
were replaced with language describing the “balance of risks” regarding the FOMC’s mandated goals of maximum employment and price stability. The August 2003 FOMC statement
said, “The Committee believes that policy accommodation can be maintained for a considerable period.” In January 2004 the forward-looking language was “the Committee believes
that it can be patient in removing its policy accommodation,” and that of May 2004 was
“policy accommodation can be removed at a pace that is likely to be measured.” As inflation
fears rose thereafter, the December 2005 statement included the words “further policy firming may be needed.”
campbell, evans, fisher, and justiniano
9
When measuring the market impact of FOMC statements, one must
confront the possibility that their content is more confirming of macro
economic conditions already known by market participants than revealing
of adjustments to policy. Failure to control for statements’ confirming content could lead to incorrectly attributing to them outcomes that are in fact
due to other factors driving revisions to expectations of growth and inflation. GSS overcome this difficulty by studying the behavior of expected
federal funds rates in symmetric 30- and 60-minute windows surrounding
the release of FOMC statements. Focusing on these narrow windows keeps
the economic information available to market participants essentially fixed.
The within-day data on which GSS rely are unavailable to us after 2004,
so we extend their work using daily observations of implied future interest
rates at the market’s close from five futures contracts: the current-month
and 3-month-ahead federal funds futures contracts (with a scale factor to
account for the timing of FOMC meetings within the month) and the 2-,
3-, and 4-quarter-ahead Eurodollar futures contracts (adjusted by the difference between the spot Eurodollar and federal funds rates); to each of
these we add a risk premium of 1 bp per month.10 Using data from the
same contracts spanning February 1990 through February 2004, GSS find
that just two factors explain more than 90 percent of the variation in these
contracts’ prices. Despite the potentially unlimited complexity of monetary
policy statements, financial markets nonetheless have reacted as if there
is essentially only one additional degree of information beyond surprise
changes in the federal funds rate target. By performing a suitable rotation
of the two factors, GSS show that they can be given “target” and “path”
interpretations. The target factor accounts for most of the surprise change
in the current federal funds rate. By construction, the path factor influences
only expected future rates.11
We begin our analysis by replicating theirs over a slightly longer time
sample, February 1990 through June 2007. We have found that many of
10. Our use of the daily window should not be too problematic, since GSS’s results are
similar when they use the daily window (see their table 1). The short windows studied by
GSS are mostly relevant for the period before February 1994, when open-market operations
were sometimes conducted following the release of labor market data on the same day.
11. GSS show that the path factor is associated with well-known significant changes
in FOMC statement language. For example, its largest realization in absolute value occurs
on January 28, 2004, when the federal funds target was not changed but the phrase “policy
accommodation can be maintained for a considerable period” was replaced with “the Committee believes it can be patient in removing its policy accommodation.” This change in
language was interpreted by markets as indicating that the FOMC would begin tightening
policy sooner than previously expected.
10
Brookings Papers on Economic Activity, Spring 2012
Table 1. Decomposing the Variance in Changes in Expected Federal Funds Rates,
1990–2007 and 1994–2007a
Percent
Share of variance due to indicated factor
Federal funds rate
futures contract
Current quarter
Next quarter
Two quarters hence
Three quarters hence
Four quarters hence
February 1990–June 2007
sample
February 1994–June 2007
sample
Target factor
Path factor
Target factor
Path factor
98
82
51
36
21
0
14
47
63
77
97
74
31
18
7
0
22
67
81
90
Source: Authors’ calculations.
a. Expected interest rates are measured using daily federal funds futures prices and Eurodollar futures
prices as described in the text. Numbers do not sum to 100 because the two factors do not explain all the
variation in the expected rate changes.
our results are sensitive to including the observation for September 2001,
so we omit it from this and all subsequent analysis in this section (as do
GSS in their online appendix). The first two columns of table 1 report the
fractions of innovation variance for each interest rate futures contract rate
that are due to the identified target factor and to the identified path factor
over this sample period. The path factor accounts for no changes to the current quarter’s interest rate by construction, and it accounts for only 14 percent of the variance in the interest rates expected for the next quarter. The
target factor accounts for nearly all of the remaining variance from these
two contracts. The path and target factors each explain about 50 percent of
the variance in interest rates expected 2 quarters hence, and the path factor
accounts for the clear majority of the variance in the two longest contracts.
Before February 1994 the FOMC did not explicitly announce changes
in its target for the federal funds rate. Although GSS show that even
before that date, market participants were able to discern within minutes
of an open-market operation whether the FOMC had changed its target,
one might reasonably suspect that little forward guidance came out of
these earlier FOMC meetings. The second two columns of table 1 report
the results when we discard these first 4 years. As expected, this change
in the sample period increases the path factor’s importance.
GSS document substantial positive statistical relationships between
their identified factors and yields on financial assets. In particular, a positive 100-bp realization of their target factor raises 2-, 5-, and 10-year
Treasury yields by 41, 37, and 28 bp, respectively (penultimate column
campbell, evans, fisher, and justiniano
11
of their table 5). Table 2 reports analogous regressions for the path and
target factors as we identify them for the two samples. (We normalize the
target factor loading on the current funds rate and the path factor loading
on the 4-quarters-ahead futures rate to be unity. GSS use a slightly different
normalization. The normalization has no impact on statistical significance
or decomposition of variance.) The table’s top panel reports the regressions
using the 2-, 5-, and 10-year Treasury yields. GSS find that the two factors
explain 94 percent, 80 percent, and 74 percent of the variance in these rate
changes, respectively. The two factors we identify have similarly strong
explanatory power for both samples we consider. For the longer sample
(first two columns), all of the slopes multiplying the factors are positive and
statistically significant at the 1 percent level. Their magnitudes are comparable to those reported by GSS, but our path factor slopes are somewhat
larger and our target factor slopes a bit smaller than theirs. For the sample
excluding the period without regular post-FOMC meeting statements (last
two columns), the target factor’s slopes are smaller and those of the path
factor larger than for the longer sample. The table’s bottom panel reports
the results using yields on Aaa/AAA- and Baa/BBB-rated corporate bonds
with at least 20 years remaining before maturity. We find these to be of
particular interest because they correspond to interest rates that are directly
relevant for firms’ investment decisions. Surprisingly to us, the target factor
has no detectable influence on these yields, regardless of which sample we
use. In contrast, a 100-bp positive path factor realization raises both yields
by about 30 to 35 bp, depending on the sample used for estimation.
Our first substantial extension of GSS uses the identified factors and
observations of private inflation and unemployment expectations to measure the macroeconomic effects of forward guidance. For this analysis we rely on the Blue Chip Economic Indicators forecast survey. At the
beginning of each month, Blue Chip solicits projections for key economic variables, including quarterly changes in the CPI and the civilian
unemployment rate, from about 50 private forecasters. From these it compiles a “consensus” forecast for each variable, which is then published
on the 10th of the month. The forecasts cover the previous quarter’s data
(which might not yet be published at the time of the survey) and each quarter in the current and next calendar years. Therefore, the data always report
a 1-quarter backcast, a current-quarter nowcast, and forecasts for at least
the next 4 quarters.12
12. The quarterly unemployment rate is expressed as the average monthly value across
the quarter’s constituent months.
0.040
(0.033)
0.051*
(0.028)
0.474***
(0.030)
0.319***
(0.043)
0.157***
(0.050)
Target factor
0.310***
(0.041)
0.313***
(0.036)
0.695***
(0.032)
0.705***
(0.041)
0.575***
(0.042)
Path factor
0.45
0.41
0.72
0.82
0.87
Adjusted R
2
-0.003
(0.046)
-0.018
(0.046)
0.287***
(0.034)
0.141**
(0.061)
-0.014
(0.073)
Target factor
0.352***
(0.045)
0.325***
(0.042)
0.698***
(0.037)
0.703***
(0.039)
0.564***
(0.037)
Path factor
0.44
0.44
0.73
0.82
0.85
Adjusted R2
February 1994–June 2007 sample
Source: Authors’ regressions.
a. Each row in each panel reports coefficients from a regression of daily changes in yields of the indicated asset on the two factors. Both samples exclude September 2001.
Robust standard errors are in parentheses. Asterisks indicate statistical significance at the *10 percent, **5 percent, and ***1 percent level.
b. Both samples include only bonds with 20 or more years to maturity.
Baa/BBB-rated
Corporate bondsb
Aaa/AAA-rated
10 years to maturity
5 years to maturity
Treasuries
2 years to maturity
Asset
February 1990–June 2007 sample
Table 2. Regressions Estimating Asset Price Responses to Target and Path Factors, 1990–2007 and 1994–2007a
campbell, evans, fisher, and justiniano
13
For each month we calculate the revisions to the forecasts of unemployment and CPI inflation for the current and next 3 quarters. Virtually
by construction, these are uncorrelated across time.13 We then regress
these revisions against the identified target and path factors. Table 3
reports the estimates (in basis points per positive 1-bp factor realization) for both precrisis samples. The first notable result is that the R2s for
these regressions are far lower than those from the analogous asset price
regressions in table 2. Since the regressions’ residuals account for all
macroeconomic news arriving in the month except that in FOMC statements, this low explanatory power is expected.
If surprise FOMC policy announcements represent shocks to the stance
of monetary policy unrelated to current macroeconomic circumstances,
then a positive innovation to either factor should raise unemployment and
lower inflation. Our estimates indicate that the opposite is more typical. For
the longer sample, the coefficients on the target factor are statistically significant and negative for unemployment expectations at all four horizons
(top panel of table 3). The path factor’s coefficients are also all negative, but
in only one case is the coefficient statistically significant (at the 10 percent
level). Switching to the shorter sample brings the estimates of the target
factor’s coefficients close to zero and amplifies the negative coefficients on
the path factor. Only 3 of the 16 estimated coefficients for inflation (bottom
panel) are negative, and none of these are statistically significant. However,
the coefficient on the path factor in the current quarter’s regression and that
on the target factor in the next quarter’s regression are significant at the
10 percent and the 5 percent level, respectively, in the later sample.
The counterintuitive signs of the estimates in table 3 require an explanation. The one we favor interprets the GSS forward guidance as Delphic:
the public believes that the FOMC has information about macroeconomic
fundamentals that the public does not, and that monetary policy surprises
arise from this informational advantage. In that case the forecast revision
following a positive policy rate innovation encompasses the revelation of
unexpectedly strong macroeconomic fundamentals as well as the contractionary effects of the innovation itself.
I.B. Forward Guidance since the Financial Crisis
The evidence that market participants and professional forecasters
are influenced by FOMC forward guidance is suggestive for the current
13. Krane (2011) searches for bias and forecast error predictability in the Blue Chip
consensus forecasts for GDP growth and finds none. Similarly, we find no evidence that the
Blue Chip forecasts of inflation and unemployment are seriously deficient.
0.25
(0.33)
0.14
(0.11)
0.11
(0.14)
0.13
(0.20)
-0.21***
(0.08)
-0.18**
(0.09)
-0.27***
(0.08)
-0.26***
(0.09)
Target factor
0.47
(0.36)
0.30
(0.24)
-0.06
(0.13)
0.07
(0.20)
-0.08
(0.06)
-0.12
(0.08)
-0.13*
(0.07)
-0.08
(0.08)
Path factor
0.01
0.01
0.03
0.02
0.07
0.09
0.05
0.07
Adjusted R
2
-0.13
(0.34)
0.25**
(0.13)
0.14
(0.10)
0.04
(0.14)
-0.01
(0.08)
0.07
(0.10)
-0.06
(0.11)
-0.03
(0.09)
Target factor
0.57*
(0.31)
0.12
(0.12)
-0.04
(0.16)
0.27
(0.25)
-0.08
(0.07)
-0.16**
(0.08)
-0.16*
(0.09)
-0.19**
(0.08)
Path factor
0.03
0.01
0.03
0.02
0.04
0.03
0.03
0.01
Adjusted R2
February 1994–June 2007 sample
Source: Authors’ regressions.
a. Each row in each panel reports coefficients from a regression of changes in monthly forecasts of either the unemployment rate or CPI inflation on the two factors. Both
samples exclude September 2001. Robust standard errors are in parentheses. Asterisks indicate statistical significance at the *10 percent, **5 percent, and ***1 percent level.
3 quarters hence
2 quarters hence
Next quarter
CPI inflation
Current quarter
3 quarters hence
2 quarters hence
Next quarter
Unemployment rate
Current quarter
Forecast
February 1990–June 2007 sample
Table 3. Regressions Estimating Private Forecast Responses to Target and Path Factors, 1990–2007 and 1994–2007a
campbell, evans, fisher, and justiniano
15
situation, but we hesitate to apply it directly to the present when the ZLB
has robbed the FOMC of its principal policy tool. Research on monetary
policy announcements since the onset of the crisis has focused almost
exclusively on the impact of the FOMC’s announcements of large-scale
asset purchases (LSAPs).14 There is significant evidence that LSAP policies can alter long-term interest rates. For example, Gagnon and others
(2010) present an event study of QE1 that documents large reductions
in interest rates concurrent with LSAP announcements. Krishnamurthy
and Vissing-Jorgensen (2011) evaluate the impact on interest rates of
announcements associated with both QE1 and QE2. They uncover several channels through which these announcements have had an impact
on asset prices and ascribe a major role to their signaling of lower future
federal funds rates. This suggests that one feature of LSAPs resembles
forward guidance, and so the findings of Krishnamurthy and VissingJorgensen (2011) can be interpreted as supporting the view that forward
guidance has significantly influenced asset prices in the recent period.
However, the recent impact of “pure” forward guidance, where the policy action is reflected solely in statement language, remains unclear.
To shed further light on the impact of forward guidance, we apply
the GSS methodology to FOMC statements issued since the onset of the
financial crisis. Table 4 presents our compilation of relevant statements
and the language in each that we judge most pertinent to forward guidance.15 The list includes the statements following every scheduled and
unscheduled FOMC meeting since August 2007 (39 in all) as well as the
November 25, 2008, Board of Governors press release that announced
the first stage of QE1. (All LSAP announcements since that press release
have been made in postmeeting FOMC statements.) Although several
remarks in speeches and testimony by Federal Reserve officials also seem
to have been interpreted by markets as forward guidance, we exclude
14. One exception is Wright (2012), who documents the effects of monetary policy
surprises on long-term interest rates since the attainment of the ZLB. His analysis draws
on identification by heteroskedasticity and does not distinguish between two factors
capturing surprises at different horizons over the expected policy path. Swanson and
Williams (2012) also discuss the effects of FOMC announcements on long-term yields,
but they focus on the responses of medium- and longer-term interest rates to macro
economic news.
15. We omit the large number of Federal Reserve press releases focused on programs
designed to promote the smooth functioning of credit markets because they did not concern
the traditional focus of countercyclical monetary policy.
16
Brookings Papers on Economic Activity, Spring 2012
Table 4. Forward Guidance in Official FOMC Statements,
August 2007–December 2011a
Date of statement
Federal
funds target
rate (%)
August 7, 2007
5.25
August 17, 2007
5.25
September 18, 2007b
4.75
October 31, 2007
4.50
December 11, 2007
4.25
January 22, 2008b
January 30, 2008
March 18, 2008
April 30, 2008
3.50
3.00
2.25
2.00
June 25, 2008
2.00
August 5, 2008
2.00
September 16, 2008
2.00
October 8, 2008b
1.50
October 29, 2008
November 25, 2008b
(press release)
1.00
0–0.25
Relevant language
“. . . the Committee’s predominant policy concern
remains the risk that inflation will fail to moderate
as expected.”
“. . . the downside risks to growth have increased
appreciably.”
“Developments in financial markets . . . have increased
the uncertainty surrounding the economic outlook.”
“. . . the upside risks to inflation roughly balance the
downside risks to growth.”
“Recent developments . . . have increased the uncertainty surrounding the outlook for economic growth
and inflation.”
“Appreciable downside risks to growth remain.”
“. . . downside risks to growth remain.”
Same as previous
“The substantial easing of monetary policy to date,
combined with ongoing measures to foster market
liquidity, should help to promote moderate growth
over time and to mitigate risks to economic activity.”
“Although downside risks to growth remain, they appear to have diminished somewhat, and the
upside risks to inflation and inflation expectations
have increased.”
“Although downside risks to growth remain, the upside
risks to inflation are also of significant concern to
the Committee.”
“The downside risks to growth and the upside risks
to inflation are both of significant concern to the
Committee.”
“Incoming economic data suggest that the pace of
economic activity has slowed markedly in recent
months. Moreover, the intensification of financial
market turmoil is likely to exert additional restraint
on spending, partly by further reducing the ability of
households and businesses to obtain credit. Inflation
has been high, but the Committee believes that the
decline in energy and other commodity prices and the
weaker prospects for economic activity have reduced
the upside risks to inflation.”
“. . . downside risks to growth remain.”
“. . . purchases [of $100 billion of GSEs and
$500 billion of MBSs] are expected to take place
over several quarters.”
17
campbell, evans, fisher, and justiniano
Table 4. Forward Guidance in Official FOMC Statements,
August 2007–December 2011a (Continued)
Date of statement
Federal
funds target
rate (%)
December 16, 2008
0–0.25
January 28, 2009
0–0.25
March 18, 2009
(QE1 announcement)
0–0.25
April 29, 2009
0–0.25
June 24, 2009
0–0.25
Relevant language
“. . . the Committee anticipates that weak economic
conditions are likely to warrant exceptionally low
levels of the federal funds rate for some time. The
focus of the Committee’s policy going forward will
be to . . . stimulate the economy through open market
operations and other measures that sustain the size of
the Federal Reserve’s balance sheet at a high level. . . .
The Committee is also evaluating the potential benefits
of purchasing longer-term Treasury securities.”
“The Committee continues to anticipate that economic
conditions are likely to warrant exceptionally low
levels of the federal funds rate for some time. The
Committee also is prepared to purchase longerterm Treasury securities if evolving circumstances
indicate that such transactions would be particularly
effective in improving conditions in private credit
markets.”
“. . . the Committee will maintain the target range for
the federal funds rate at 0 to ¼ percent and anticipates
that economic conditions are likely to warrant exceptionally low levels of the federal funds rate for an
extended period. The Committee sees some risk that
inflation could persist for a time below rates that
best foster economic growth and price stability in
the longer term. . . . The Committee decided today
to increase the size of the Federal Reserve’s balance
sheet further by purchasing up to an additional
$750 billion of [MBSs], bringing its total purchases of
these securities to up to $1.25 trillion this year, and to
increase its purchases of [GSE] debt this year by up to
$100 billion to a total of up to $200 billion. . . . The
Committee decided to purchase up to $300 billion
of longer-term Treasury securities over the next six
months.”
“. . . Committee sees some risk that inflation could persist for a time below rates that best foster economic
growth and price stability in the longer term. . . .
Economic conditions are likely to warrant exceptionally low levels of the federal funds rate for an
extended period.”
“. . . economic conditions are likely to warrant exceptionally low levels of the federal funds rate for an
extended period. . . . The Committee expects that
inflation will remain subdued for some time.”
(continued)
Table 4. Forward Guidance in Official FOMC Statements,
August 2007–December 2011a (Continued)
Date of statement
Federal
funds target
rate (%)
August 12, 2009
0–0.25
September 23, 2009
0–0.25
November 4, 2009
0–0.25
December 16, 2009
0–0.25
January 27, 2010
March 16, 2010
April 28, 2010
June 23, 2010
August 10, 2010
0–0.25
0–0.25
0–0.25
0–0.25
0–0.25
September 21, 2010
0–0.25
November 3,
2010 (QE2
announcement)
0–0.25
December 14, 2010
January 26, 2011
March 15, 2011
April 27, 2011
June 22, 2011
August 9, 2011
0–0.25
0–0.25
0–0.25
0–0.25
0–0.25
0–0.25
September 21, 2011
November 2, 2011
December 13, 2011
0–0.25
0–0.25
0–0.25
Relevant language
“Although economic activity is likely to remain weak
for a time, the Committee continues to anticipate that
policy actions to stabilize financial markets and institutions, fiscal and monetary stimulus, and market forces
will contribute to a gradual resumption of sustainable
economic growth in a context of price stability. . . .
Substantial resource slack is likely to dampen cost
pressures, and the Committee expects that inflation
will remain subdued for some time.”
“. . . economic conditions are likely to warrant exceptionally low levels of the federal funds rate for an
extended period. . . . [MBS and GSE purchases will
finish by the] end of the first quarter of 2010.”
“. . . economic conditions . . . are likely to warrant
exceptionally low levels of the federal funds rate for
an extended period [and the Committee will complete purchases of GSE debt of about $175 billion].”
“. . . economic conditions . . . are likely to warrant
exceptionally low levels of the federal funds rate for
an extended period.”
Same as previous
Same as previous
Same as previous
Same as previous
Same as previous, plus “the Committee will keep constant the Federal Reserve’s holdings of securities at
their current level by reinvesting principal payments
from agency debt and agency [MBSs] in longer-term
Treasury securities.”
Same as June 23, plus “The Committee also will
maintain its existing policy of reinvesting principal
payments from its securities holdings.”
Same as previous, plus “In addition, the Committee
intends to purchase a further $600 billion of longerterm Treasury securities by the end of the second
quarter of 2011.”
Same as previous
Same as previous
Same as previous
Same as previous
Same as previous
“. . . economic conditions . . . are likely to warrant
exceptionally low levels of the federal funds rate at
least through mid-2013.”
Same as previous
Same as previous
Same as previous
Source: Board of Governors of the Federal Reserve System website at www.federalreserve.gov/news
events/press/monetary/2012monetary.htm.
a. The November 28, 2008, press release was issued by the Board of Governors of the Federal Reserve
System. All other statements were issued by the FOMC. GSE = government-sponsored enterprise; MBS =
mortgage-backed security.
b. Statement was issued between regularly scheduled FOMC meetings.
campbell, evans, fisher, and justiniano
19
these from our analysis, since it is difficult to find an objective criterion
for including any given instance.16
Mimicking our analysis of the precrisis period, we estimate factors from
changes in expected future federal funds rates between the close of business
the day before and the day of each of the announcements listed in table 4.
Because the horizon over which forward guidance is issued seems to be
longer since the crisis than it was during the precrisis period, we examine the behavior of seven futures contracts that pin down the expected
path of the federal funds rate over the next year and a half without overlapping: the current-month and 3-month-ahead federal funds futures
contracts (again with a scale factor to account for the timing of FOMC
meetings within the month) and the 2-, 3-, 4-, 5-, and 6-quarter-ahead
Eurodollar futures contracts (again adjusted by the difference between
the spot Eurodollar and federal funds rates). As before, we also adjust
all rates for an assumed risk premium of 1 bp per month. Just as in the
precrisis period, two factors explain most of the variability in the futures
data. Henceforth we focus on the first two factors after they have been
rotated as in GSS.
Figure 1 is a scatterplot of the path factor against changes in the 10-year
Treasury yield for the 40 dates listed in table 4. We distinguish statements
containing announcements of LSAPs from other statements, and the statements most closely associated with QE1 and QE2 (March 18, 2009, and
November 3, 2010, respectively) are labeled. The most striking feature of
figure 1 is how much of an outlier the March 18, 2009, announcement is. On
that date the 10-year yield fell (as intended) 51 bp while the path factor rose
32 bp. Markets interpreted the FOMC’s announcement as indicating that the
recovery would come sooner than previously thought and that, consequently,
liftoff in the federal funds rate from the ZLB would come earlier than previously anticipated; the 2-quarter-ahead futures contract rose 60 bp from the
day before. In contrast, the response to the QE2 announcement appears very
much like the responses to the other FOMC announcements, which indicate
a positive relationship between the path factor and changes in the 10-year yield.
Indeed, Krishnamurthy and Vissing-Jorgensen (2011, p. 217) find that “the
16. Probably the most relevant instances in this regard are speeches on December 1,
2008, and August 27, 2010, by Federal Reserve Chairman Ben Bernanke, which were interpreted by markets as opening the door to the first and second round of large-scale purchases
of Treasury securities, respectively. With the exception of the December 1, 2008, speech,
our compilation includes every QE1 and QE2 date employed in Krishnamurthy and VissingJorgensen’s (2011) event study.
20
Brookings Papers on Economic Activity, Spring 2012
Figure 1. Path Factor and Changes in 10-Year Treasury Yields on FOMC Statement Dates
LSAP Announcements
Other FOMC Statements
40
QE1 3/18/2009
Path factor
20
QE2 11/3/2010
0
–20
–40
–60
–40
–20
Change in 10-Year Note
0
20
Source: Haver Analytics/Federal Reserve H.15 and authors’ calculations based on Chicago Mercantile
Exchange data.
main effect on corporate bonds and [mortgage-backed securities] in QE2
appears to have been through a signaling channel, whereby financial markets
interpreted QE as signaling lower federal funds rates going forward.”
The apparently very different response to the March 18, 2009, QE1
announcement motivates us to exclude it from the remainder of our factor analysis.
Table 5 reports the fractions of variance in changes to expected future
federal funds rates explained by the target and by the path factor estimated
from all the announcements in table 4 except the outlier associated with
QE1. The target factor dominates the variation in the current-quarter rate
and the 1-, 2-, and 3-quarter-ahead rates, whereas the path factor explains
the majority of variation in the three longer rates and negligible shares of
the three shortest contracts. This pattern is broadly similar to that for the
precrisis period reported in table 1. The main difference is that here the
path factor dominates only those changes in expected interest rates that are
4 or more quarters ahead.
Table 6 reports asset price regression estimates analogous to those of
table 2, based on the postcrisis factors. Since this sample is smaller, the
estimates’ associated standard errors are larger. These estimates strongly
21
campbell, evans, fisher, and justiniano
Table 5. Decomposing the Variance in Changes in Expected Federal Funds Rates,
August 2007–December 2011a
Percent
Share of variance due to
indicated factor
Federal funds rate futures contract
Current quarter
Next quarter
Two quarters hence
Three quarters hence
Four quarters hence
Five quarters hence
Six quarters hence
Target factor
Path factor
94
98
93
57
44
31
16
0
0
3
35
53
68
79
Source: Authors’ calculations.
a. Expected interest rates are measured using daily federal funds futures prices and Eurodollar futures
prices as described in the text. Numbers do not sum to 100 because the two factors do not explain all the
variation in the expected rate changes.
Table 6. Regressions Estimating Asset Price Responses to Target and Path Factors,
August 2007–December 2011a
Asset
Treasuries
2 years to maturity
5 years to maturity
10 years to maturity
Corporate bondsb
Aaa/AAA-rated
Baa/BBB-rated
Target factor
Path factor
Adjusted R2
0.592***
(0.096)
0.404***
(0.143)
0.250*
(0.131)
0.716***
(0.160)
0.898***
(0.165)
0.877***
(0.103)
0.79
0.058
(0.079)
0.065
(0.085)
0.631***
(0.085)
0.556***
(0.117)
0.66
0.58
0.45
0.34
Source: Authors’ regressions.
a. Each row in each panel reports coefficients from a regression of daily changes in yields of the indicated
asset on the two factors. Robust standard errors are in parentheses. Asterisks indicate statistical significance
at the *10 percent, **5 percent, and ***1 percent level.
b. Both samples include only bonds with 20 or more years to maturity.
resemble those from before the crisis. Both factors have a large positive
influence on the 2- and 5-year yields, and the path factor substantially influences the 10-year Treasury yield and yields on seasoned Aaa/AAA- and
Baa/BBB-rated corporate bonds. Given the disparity in economic conditions
between the pre- and postcrisis sample periods, the similarity of forward
guidance effects on asset prices is a striking finding.
22
Brookings Papers on Economic Activity, Spring 2012
Table 7. Regressions Estimating Private Forecast Responses to Target and Path Factors,
August 2007–December 2011a
Forecast
Unemployment rate
Current quarter
Next quarter
2 quarters hence
3 quarters hence
CPI inflation
Current quarter
Next quarter
2 quarters hence
3 quarters hence
Target factor
Path factor
Adjusted R2
–0.21
(0.19)
–0.29
(0.26)
–0.33
(0.34)
–0.35
(0.39)
0.01
(0.31)
0.02
(0.47)
0.11
(0.62)
0.15
(0.73)
0.02
1.80
(1.82)
0.53
(0.64)
–0.01
(0.12)
0.07
(0.11)
2.05
(4.17)
0.44
(1.43)
–0.02
(0.27)
0.23
(0.29)
0.07
0.03
0.04
0.03
0.04
0.00
0.03
Source: Authors’ regressions.
a. Each row in each panel reports coefficients from a regression of changes in monthly forecasts of either
the unemployment rate or CPI inflation on the two factors. Robust standard errors are in parentheses.
Table 7 reports estimates for the forecast innovation regressions using
the postcrisis data. The estimated standard errors greatly exceed those
from the analogous regressions estimated with precrisis data (table 3),
so that none of the reported coefficients are statistically significant.
Although we conclude that our regression estimates of the effects of
forward guidance on macroeconomic expectations since the financial
crisis are too imprecise to allow strong quantitative conclusions, the
estimates are broadly consistent with those from the precrisis period.
II. Forward Guidance through an Interest Rate Rule
The event-study approach used above isolates “pure” forward guidance
associated with distinct policy announcements from other monetary policy
actions, but it fails to identify any forward guidance communicated through
other channels. In this section we present a new and complementary methodology that identifies forward guidance communicated through all the
channels available to the FOMC. This approach builds on the long-standing
campbell, evans, fisher, and justiniano
23
practice of summarizing monetary policy with a parsimonious rule for setting
the policy rate. By applying such a rule both to actual policy decisions and
to observations of private expectations, we are able to identify consensus
expectations of how the FOMC will deviate from the monetary policy rule
at a specific date in the future.
The empirical implementation of our methodology inserts the Blue
Chip forecasts and interest rate futures prices examined above, aggregated to quarterly frequency, into an interest rate rule with two lags of
the interest rate and measures of the unemployment gap and inflation.
The rule’s novelty lies in its residual, which sums components gradually revealed to the public up to 4 quarters before the policy action. The
interest rate futures and professional forecasts together are sufficient to
identify these forward guidance shocks. For the period 1996Q1 through
2007Q2, the estimated rule describes the 4-quarter-ahead expectation
of the interest rate very well: the standard deviation of the 4-quarterahead forward guidance shock is only 9 bp. The standard deviation of
the interest rate rule’s total residual (which sums the forward guidance
shocks with a traditional unanticipated policy shock) is 30 bp. However,
the standard deviation of the anticipated component is 28 bp. That is, the
Federal Reserve successfully telegraphs most departures from the interest rate rule in advance.
The forward guidance shocks we identify from the interest rate rule
differ from the statement date–based shocks of GSS in some ways and
resemble them in others. The most notable difference is their factor
structure. The contemporaneous policy shock and the four forward guidance shocks revealed every quarter have a single factor that explains
most of the 4-quarter-ahead forward guidance but much less at closer
horizons. A positive realization of this factor speeds up the usual interest rate changes following a contemporaneous monetary policy shock,
so we call it the policy acceleration factor. The FOMC seems to have
used this factor heavily during the 2001 recession and in its aftermath.
The similarities between GSS-style forward guidance shocks and those
measured with an interest rate rule become apparent when we calculate
their effects on asset prices and macroeconomic forecasts. Positive forward guidance shocks raise both Treasury and corporate bond yields.
By construction, the interest rate rule accounts for the FOMC’s typical responses to varying economic fundamentals as measured by inflation and the unemployment gap. Nevertheless, regressions analogous
to those in table 3 indicate that the same anticipated deviations from
24
Brookings Papers on Economic Activity, Spring 2012
this rule affect unemployment and inflation forecasts with the “wrong”
sign, just as do the statement date–based GSS shocks. We interpret these
results as arising from the FOMC adjusting policy quickly when revisions
to macroeconomic expectations catch it “behind the curve.”
II.A. Rule-Based Measurement of Forward Guidance
We consider interest rate rules for the average policy rate over quarter t,
rt, of the following form:
(1)
rt = µ + ρ1rt −1 + ρ2 rt − 2
M
+ (1 − ρ1 − ρ2 ) ( φ π π t + φu µ t ) + ∑ νt − j , j .
j=0
The variables p~t and u~t are the policy-relevant measures of the inflation rate
and the unemployment gap (the difference between the unemployment rate
and a measure of the economy’s non-accelerating-inflation, or “natural,”
unemployment rate). Parameters r1, r2, fp, and fu determine the degree of
interest smoothing and how the policy rate responds to typical changes in
macroeconomic conditions.
The distinguishing feature of equation 1 is the last term, which involves
the M + 1 disturbances, nt-j,j for j = 0, 1, . . . , M. The first of these, nt,0, is
the monetary policy disturbance that appears in conventional interest rate
rules. It captures the Federal Reserve’s response to extraordinary events,
such as the September 11 terrorist attacks or the 1997 Asian currency crisis, that warrant a rapid but temporary deviation from the normal policy
prescription. The remaining disturbances are forward guidance shocks,
because they are revealed to the public before they are applied to the interest rate rule. The public sees nt,j in quarter t, and the FOMC applies it to
the rule j quarters hence. We gather all of the shocks revealed in quarter t
into the vector nt ≡ (nt,0, nt,1, . . . , nt,M). Each realization of nt influences the
expected path of interest rates. To identify the forward guidance shocks, we
wish to map revisions to expectations, which are uncorrelated over time by
construction, onto realizations of nt; so we assume that nt is also uncorrelated over time. That is, we assume that the elements of nt are news relative
to the public information set at the end of t - 1. For sufficiently large M
and under rational expectations, this can be done without loss of generality.17 Although nt is uncorrelated over time, its elements may be correlated
17. The reason is that a time-series variable at time t always can be decomposed into the
sum of its expected value based on information available at t - 1 and an orthogonal innovation.
campbell, evans, fisher, and justiniano
25
with each other. Allowing for this correlation admits the possibility that the
FOMC provides information on multiple future quarters’ monetary policy
shocks in the same communication.
The practice of including exogenous shocks to the interest rate is commonplace. Our specification differs from conventional interest rate rules
only in the assumption that the public observes some of the interest rate
shocks before their implementation. The most similar recent work is that of
Stefan Laséen and Lars Svensson (2011), who propose modeling forward
guidance with an interest rate rule as we do when calculating the equilibrium of a New Keynesian model.
One can recover nt using data on private expectations of unemployment,
inflation, and the federal funds rate with values of r1, r2, fp, and fy in hand.
Here and henceforth, conditional expectations at quarter t are defined in
terms of information at the beginning of the quarter.18 For any variable x,
we denote its realization in quarter t with xt. Then we use the notation x tj
to denote the time t - j conditional expectation of variable xt. Since not all
variables dated t are known by economic agents at the start of the quarter
they are realized, the “nowcast” x t0 does not necessarily equal the realized
xt. For example, r t0 is the expectation at the beginning of quarter t of the
quarter’s average policy rate, which can clearly change over the quarter.
If x is not even revealed to the public during the quarter of its realization,
then the “backcast” xt-1 also might not equal xt. The unemployment rate
provides a relevant example. Its backcast differs from its realized value
because the time taken for its tabulation delays its release.
To measure nt-M,M, suppose that the public expects the FOMC to follow
equation 1 on average. Then, taking expectations given information at the
start of period t - M + 1 yields
(2)
rt M −1 = µ + ρ1rt −M1− 2 + ρ2rt −M2− 3
+ (1 − ρ1 − ρ2 )( φπ π tM −1 + φu u˜ tM −1 ) + νt − M, M .
The residual term in equation 2 equals nt-M,M because the expected value
Et-M+1[nt,j] = 0 for j = 0, . . . , M - 1. Thus, nt-M,M equals the deviation of
the expected interest rate M - 1 quarters ahead from its value dictated by
the interest rate rule’s expected value. To recover the other errors, we take
18. This conforms to the timing convention used for the Blue Chip macroeconomic
expectations data.
26
Brookings Papers on Economic Activity, Spring 2012
expectations of equation 1 at two adjacent dates and difference the results.
For 0 ≤ j < M we obtain
(3)
r tj −1 − rt j = ρ1 ( r tj −−12 − r tj −−11 ) + ρ2 ( r tj −− 32 − r tj −− 22 )
+ (1 − ρ1 − ρ2 )[ φ π ( π tj −1 − π tj ) + φu ( u tj −1 − u tj )] + νi − j, j .
Equation 3 shows that nt-j,j equals the change within quarter t - j in the
expected interest rate for quarter t corrected for the change in the interest
rate rule’s expected value arising from revisions in private expectations
of inflation and unemployment. This disturbance embodies expected
deviations from “typical” monetary policy. Forward guidance influences
nt-j,j when the FOMC communicates a prospective change in its shortrun policy goals with or without a credible Odyssean commitment. The
anticipated residuals might also arise from external factors omitted from
the rule, but only to the extent that they affect the policy rate through
channels other than the forecasts of the unemployment gap and inflation
that already appear in the rule. How much weight is given to a conditioning variable when constructing a forecast depends on the prevailing
economic conditions. For example, before the increase in foreign trade
associated with globalization, there was less need to pay attention to
foreign inflation and the exchange rate than there is today. This does not
necessarily mean that the policy rule incorrectly omits foreign inflation
or the exchange rate, because these variables are an input into agents’
forecasts.
II.B. Estimation
Implementing this methodology requires observations of private expectations and the estimation of µ, r1, r2, fp, and fu. The Blue Chip consensus
0
0
-1
j
forecasts give us u -1
t-1 and p t-1 (backcasts), u t and p t (nowcasts), and u t+j and
j
p t+j for j = 1, . . . , 4 (forecasts). In March and October, Blue Chip survey
participants report forecasts of each variable’s average value 7 to 11 years
after the current calendar year. We use the most recently published consensus long-run forecast for the unemployment rate as a measure of each
quarter’s natural rate of unemployment, u*.
From this we construct the
t
expected unemployment gap in quarter t + j as û tj ≡ u tj - u*.
t Our Blue Chip
data contain observations for the period 1989Q2 through 2011Q4.
Our implementation of the interest rate rule employs averages of the
expected unemployment gap and expected inflation over the previous, cur-
campbell, evans, fisher, and justiniano
27
rent, and next quarters as perceived at the beginning of the next quarter.
That is:
u t =
1 1 j −1
∑ uˆt + j
3 j =−1
π t =
1 1 j −1
∑ πt+ j .
3 j =−1
Here we have abused our notation by supposing that u~t and p~t are realized
at the end of quarter t even though they depend on information available
“at the beginning” of quarter t + 1. We can construct forecasts of u~t and p~t
from the Blue Chip data up to 3 quarters ahead, so we set M in equation 1
equal to 4. That is, we assume that the process of communicating forward
guidance begins 4 quarters before the policy decision in question.
Although the Blue Chip data contain forecasts of the federal funds rate,
we prefer to base our measures of expected interest rates on the futures
market prices used in section I from each quarter’s final trading day. Our
estimation uses only data from the period in which federal funds futures
have been actively traded in large volume, which James Hamilton and
others (2011) identify as beginning sometime in 1994. Because the estimation requires lags, we begin our sample with the forecasts of interest rates
that prevailed in 1996Q1.19 These prices give us the interest rates that our
procedure requires when M equals 4: rt0, rt1, . . . , rt5. The other observations
required to calculate nt are u~ t0, . . . , u~ t3 and p~ t0, . . . , p~ t3. We can calculate
these with the backcast, nowcast, and four quarterly forecasts in the Blue
Chip data.
One frequent approach to estimating the parameters of an interest rate rule
simply assumes that the autoregressive terms in equation 1 sufficiently capture the interest rate’s serial correlation, so that the policy shock is serially
uncorrelated and ordinary least squares estimation can be employed. This
assumption fails if past forward guidance influences the unemployment gap
and inflation, so we require an alternative estimator. We turn to a generalized
19. Beginning the sample in 1996Q1 also excludes an outlying observation from the
Eurodollar futures market in 1994Q4 from our analysis. In that quarter the Eurodollar rate
for delivery in 1995Q4 (averaged across that quarter’s months) rose from 6.7 percent to
8.0 percent. However, it had returned to 6.5 percent by the end of 1995Q1. Such large
changes in expected future interest rates were common in the early 1990s but occurred much
less frequently in our sample period.
28
Brookings Papers on Economic Activity, Spring 2012
method of moments (GMM) implementation of an instrumental variables
strategy. From the Blue Chip data we can calculate û tM and p tM. These, r Mt-2-2,
and r Mt-1-1 are valid instruments for nt0, nt-1,1, . . . , nt-M,M because those monetary policy shocks are all revealed after the beginning of quarter t - M.
Therefore, we can construct a valid GMM estimator based on the population moment conditions
E [ gt ( γ ) ⊗ Z t ] = 0.
Here, g = (µ, r1, r2, fp, fu) is the parameter vector, gt(?) is a function that
takes the parameter values and returns the vector (nt0, nt-1,1, . . . , nt-M,M),
and Zt = (û tM, p tM, r Mt-2-2, r Mt-1-1) is the vector of instruments. With M = 4, this
provides 16 moment restrictions to estimate 4 parameters.
This moment condition underlying our GMM estimator depends on the
assumption that our interest rate rule omits no relevant information known
in quarter t - M. This assumption would be violated if the FOMC gave
forward guidance more than 4 quarters in advance. In that case the value of
nt,4 inferred using the interest rate rule’s correct parameter values should be
correlated with the instruments in Zt. The “considerable period” language
provides one obvious potential example of such long-term forward guidance. The relevant part of the August 12, 2003, statement that introduced
it reads
The Committee judges that, on balance, the risk of inflation becoming undesirably low is likely to be the predominant concern for the foreseeable future. In
these circumstances, the Committee believes that policy accommodation can be
maintained for a considerable period.
The statement’s emphasis on anticipated inflation leads us to read this as
Delphic rather than Odyssean, so we expect it to have operated through the
interest rate rule rather than through its residuals. We can think of no other
concrete examples of long-term forward guidance of any sort during our
sample period, so we believe any biases from choosing M to conform with
the Blue Chip forecast horizon to be small.20
As noted above, our estimation sample begins in 1996Q1. We consider the crisis period that arguably began in 2007Q3 to be unique, and
20. A violation of our moment condition could also arise from mismeasurement of private expectations. If the Blue Chip survey measures equal the public’s true expectations
summed with a classical measurement error, then the measurement errors contribute to g(g).
This biases our GMM estimator only to the extent that the same errors influence the measured values of û t4 and p t4 in Zt.
campbell, evans, fisher, and justiniano
29
so we end our estimation sample with 2007Q2. The estimated interest
rate rule is
rt = −0.05 + 1.60 × rt −1 − 0.66 × rt − 2
(0.02) (0.02)
(0.02)
− (1 − 0.94 ) × 1.10 u t + (1 − 0.94 )
(0.28)
4
× 2.32 π t + ∑ νt − j , j .
j=0
(0.18)
Heteroskedasticity- and autocorrelation-consistent standard errors appear
below each estimate in parentheses. The estimates’ associated J statistic is
very small (0.25), so the estimates clearly pass the test of overidentifying
restrictions.
Two features of the interest rate rule are worth noting. First, we find an
important role for second-order autoregressive dynamics. This gives the
interest rate’s response to a one-time innovation (holding u~ t and p~ t fixed)
a hump shape: monetary policy adjustments start small, grow, and persist.
Second, the estimated rule satisfies the Taylor principle that the long-run
interest rate rises more than one for one with a persistent increase in inflation. The standard error on this coefficient is small enough to comfortably
exclude the possibility that this arises only from sampling error.
II.C. How Well Does the Public Forecast Deviations
from the Interest Rate Rule?
Given the estimated parameter values, we follow the procedure presented above to recover the history of nt from the available data. The standard deviations of the forward guidance shocks by horizon are 12, 20, 13,
11, and 9 bp for nt,0 through nt,4, respectively. As noted above, the fact that
the 4-quarter-ahead forward guidance shock nt,4 has such a small standard
deviation suggests that the estimated rule summarizes medium-run expectations of the federal funds rate very well. We can use these estimates to
calculate a variance decomposition of the interest rate rule’s intercept.21
Overall, it appears that the FOMC communicates about 40 percent of the
monetary policy variance in the quarter before its realization and another
40 percent in the 1 to 3 quarters before then.
21. Although the elements of nt are correlated with each other, we assume that its realizations are independent over time. Therefore, the five shocks contributing to the interest rate
rule’s intercept in a given quarter are mutually independent.
30
Brookings Papers on Economic Activity, Spring 2012
Figure 2. The Interest Rate Rule’s Residual and Its Forward Guidance Component,
1996–2007a
Basis points
40
Interest rate rule’s residual
20
0
–20
–40
Residual’s anticipated component
–60
–80
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
Source: Authors’ calculations.
a. Shading indicates the 2001 recession.
Figure 2 gives a visual perspective on this decomposition. It plots the
composite residual for the interest rate rule Σ4j=0 nt-j,j as well as its forward
guidance component, which simply drops the contemporaneous shock nt,0.
Overall, the two series track each other quite closely. Indeed, their sample
correlation is 0.9. At the onset of the 2001 recession, however, the two
series differ by 62 bp, reflecting the well-known sudden reversal of the
monetary policy stance at that date. In the second quarter of 2001, the difference is 37 bp. Two events that do not show up with particularly large
values of nt,0 are the Asian financial crisis and September 11. The estimate
of n1997Q3,0 is only -0.8 bp. It turns out that markets anticipated during the
previous quarter most of the monetary policy accommodation provided in
that quarter. Following September 11, the FOMC increased accommodation only in 2001Q4, because the Federal Reserve concentrated on maintaining the orderly functioning of financial markets in the final weeks of
2001Q3. Nevertheless, market participants anticipated this move, so it
shows up in n2001Q3,1, estimated at -85 bp.
Since each realization of nt moves the entire expected path of interest
rates, it is reasonable to suppose that its elements correlate with each other.
Indeed, such correlation underlies the factor analysis of GSS. The sample
correlation matrix is as follows:
31
campbell, evans, fisher, and justiniano
νt, 0
νt, 1
νt, 1
ν t, 2
ν t, 3
0.02
νt, 2 −0.05
0.22
νt, 3 −0.32
0.03
νt, 4 −0.32
−0.26
−0.17
0.22 0.16
The fact that nt,0 is negatively correlated with both nt,3 and nt,4 suggests that
the public expects some “last-minute” monetary policy adjustments to be
reversed in the relatively near future. The other forward guidance shocks
are uncorrelated with nt,0, and they display relatively low correlations with
each other.
II.D. Factor Analysis
Although the correlations among the five shocks contributing to the
interest rate rule’s intercept are not large, GSS’s successful use of factor
analysis motivates us to investigate how a factor model explains them. The
negative correlations of nt,0 with nt,3 and nt,4 hint at a single factor structure
in which the factor “tilts” the monetary policy shocks, providing accommodation today while promising to take it away later. We investigate this
impression by estimating
vt = Λft + et .
Here L is a 5 × 1 matrix of factor loadings, ft is a scalar factor with a mean
of zero and variance of 1, and et is a 5 × 1 vector of mutually independent
“idiosyncratic” errors.
The maximum-likelihood estimates of the factor loadings from this
model are 5, 6, 4, -3, and -7 basis points for nt,0 through nt,4, respectively.
These estimates reveal that the factor does indeed tilt the path of monetary
accommodation: a 1-standard-deviation negative realization lowers the
interest rate rule’s intercept by about 5 bp for each of the next 3 quarters
and increases it by about the same amount for the following 2 quarters.
The factor model’s remaining parameters describe the standard deviations
of the idiosyncratic errors in et. These estimates—11, 19, 13, 10, and 6 for
nt,0 through nt,4, respectively—show that the factor accounts for about
15 percent of the variance of nt,0, about 10 percent of the variance of nt,1,
nt,2, and nt,3, and about 60 percent of the variance of nt,4. That is, the factor
32
Brookings Papers on Economic Activity, Spring 2012
Figure 3. Direct Effects of Monetary Policy Shocks on the Interest Rate
Basis points
–5
Standard contemporaneous impulse
–10
Factor impulse
–15
–20
1
2
3
4
5
6
Quarters after initial impulse
7
8
Source: Authors’ calculations.
accounts for most of 4-quarter-ahead forward guidance but leaves most
forward guidance issued at shorter horizons unexplained.
Figure 3 plots the direct interest rate effects (that is, omitting any possible endogenous responses of inflation or unemployment) over 9 quarters
of a 1-standard-deviation shock to the factor. For comparison, we also plot
the response to a standard contemporaneous impulse that initially lowers
the interest rate by the same amount (5 bp). As dictated by the secondorder autoregressive parameters, the interest rate falls for 3 quarters after
the standard contemporaneous impulse and then begins a slow rise back
to its mean. The interest rate also falls for 3 quarters following the factor
shock, but it falls much more relative to the initial response. Thereafter
the impulse’s effects dissipate quickly: after 9 quarters the interest rate has
returned to its mean. To us, these responses suggest labeling this factor
“policy acceleration.” When the factor equals zero, policy adjustments proceed at their normal pace. A negative realization increases the speed of the
interest rate’s decline and recovery, whereas positive realizations increase
the speed of impact of contractionary policy.
Figure 4 plots over time the identified policy acceleration factor scaled
by its impact on the current interest rate. This measure achieved its maximum value of 9 bp in 1999Q2, although its value in the next quarter almost
33
campbell, evans, fisher, and justiniano
Figure 4. Estimated Policy Acceleration Factor, 1996–2007
Basis points
6
3
0
–3
–6
–9
–12
–15
–18
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
Source: Authors’ calculations.
a. Shading indicates the 2001 recession.
exactly offset this promised accelerated stimulus. Its minimum of -21 bp
occurred in the wake of the 2001 recession, in 2002Q2. In that quarter the
1-, 2-, and 3-quarter-ahead forecasts of the unemployment rate all rose
30 bp. (For a point of comparison, these revisions’ sample standard errors
are 17, 20, and 21 bp, respectively.) Its other large and negative realizations
occurred during the 2001 recession itself, when the upward unemployment
forecast revisions were even larger. It appears that the FOMC successfully signaled its intention to accelerate accommodation following adverse
unemployment news in 2001 and 2002.
II.E. Asset Price and Forecast Responses to Forward Guidance Shocks
Identified from an Interest Rate Rule
One clear virtue of the GSS path factor is its documented impacts on
asset prices that are relevant for private decisions. We now examine the
impact on asset prices of the forward guidance shocks identified from the
interest rate rule by regressing the same financial variables used in table 2 on
them. Since our data are quarterly, we measure bond yields and the stock
market index on the quarter’s final trading day. The changes in these from
the previous quarter are our dependent variables. For independent variables
we use a constant and all five of the n shocks. Table 8 reports the estimated
34
Brookings Papers on Economic Activity, Spring 2012
Table 8. Regressions Estimating Asset Price Responses to Forward Guidance Shocks
Identified from an Interest Rate Rule, 1996Q1–2007Q2a
Shock
Asset
Treasuries
2 years to maturity
5 years to maturity
10 years to maturity
Corporate bondsb
Aaa/AAA-rated
Baa/BBB-rated
Constant
nt,0
nt,1
nt,2
nt,3
nt,4
5.90
(4.47)
3.46
(4.31)
1.57
(4.44)
1.08***
(0.37)
0.61*
(0.36)
0.38
(0.37)
1.98***
(0.22)
1.83***
(0.21)
1.48***
(0.22)
1.56***
(0.33)
1.91***
(0.32)
1.60***
(0.33)
0.70*
(0.42)
1.43***
(0.40)
1.41***
(0.42)
0.89*
(0.50)
1.25**
(0.49)
1.29***
(0.50)
0.60
(4.63)
0.57
(4.01)
0.19
(0.38)
0.13
(0.33)
0.65***
(0.23)
0.69***
(0.20)
0.75**
(0.34)
0.71**
(0.30)
0.86**
(0.43)
1.00***
(0.38)
0.17
(0.52)
0.37
(0.45)
Adjusted
R2
0.77
0.78
0.70
0.33
0.42
Source: Authors’ regressions.
a. Each row reports coefficients from a regression of changes in yields of the indicated asset from the
last trading day of a quarter to that of the next on a constant and on shocks nt,0 through nt,4, where nt,0 is the
monetary policy shock that occurs contemporaneously with announcement t, and the remaining shocks
nt,j are forward guidance shocks indicating the change in monetary policy announced at t to occur in quarter j. The regression coefficients can be interpreted as the response (in basis points) of the indicated asset
price to a 1-basis-point change in the indicated nt,j. Standard errors are in parentheses. Asterisks indicate
statistical significance at the *10 percent, **5 percent, and ***1 percent level.
b. Both samples include only bonds with 20 or more years to maturity.
coefficients, their standard errors, and the regressions’ R2s. We express all
of the variables in bp, so the coefficients can be read as the response in
basis points to a 1-bp change in the right-hand-side variable.
Although the coefficients’ standard errors are not small, the regression estimates clearly show that the identified forward guidance shocks
are associated with substantial changes in asset prices. A 100-bp increase
in nt,1 raises the 2- and 5-year Treasury yields by almost 200 bp and the
10-year Treasury yield by about 150 bp. The effects on the two corporate
bonds are more modest, 65 and 69 bp. In light of the standard errors, we
judge the estimated effects of nt,2 and nt,3 on these bond yields to be about
the same. The relatively small variance of nt,4 translates into relatively large
standard errors for its estimated effects on bond yields. Nevertheless, the
point estimates for the effects of nt,4 are statistically significant for the 5and 10-year Treasury yields. Overall, the estimated asset price effects
of forward guidance inferred from the interest rate rule are much larger
than the corresponding effects of forward guidance identified from the GSS
event-study methodology.
35
campbell, evans, fisher, and justiniano
Table 9. Regressions Estimating Forecast Revisions in Response to Forward Guidance
Identified from an Interest Rate Rule, 1996Q1–2007Q2a
Shock
Change in forecast
b
Constant
Unemployment rate
-6.82***
ut-1 - u t0
(2.47)
-4.02
u 0t+1 - u1t+1
(2.92)
-3.39
u1t+2 - u 2t+2
(2.93)
-2.86
u 2t+3 - u 3t+3
(2.65)
Inflation
1.83
p t-1 - p 0t
(5.55)
-5.20*
p 0t+1 - p1t+1
(2.91)
-7.55***
p1t+2 - p2t+2
(2.69)
-5.32**
p 2t+3 - p3t+3
(2.11)
Adjusted
R2
nt,0
nt,1
nt,2
nt,3
nt,4
-0.37*
(0.20)
-0.34
(0.24)
-0.46*
(0.24)
-0.31
(0.22)
-0.20
(0.12)
-0.30**
(0.14)
-0.47***
(0.14)
-0.47***
(0.13)
-0.13
(0.18)
-0.05
(0.22)
-0.02
(0.22)
-0.00
(0.20)
-0.38
(0.23)
-0.27
(0.27)
-0.20
(0.27)
-0.07
(0.25)
0.46
(0.28)
0.54
(0.33)
0.30
(0.33)
0.26
(0.30)
0.28
-0.35
(0.46)
-0.18
(0.24)
-0.05
(0.22)
-0.25
(0.18)
0.23
(0.27)
0.17
(0.14)
0.15
(0.13)
0.18*
(0.10)
-0.08 -0.61 -0.09
(0.41) (0.52) (0.63)
0.05 -0.44
0.07
(0.21) (0.27) (0.33)
0.11
0.35 -0.02
(0.20) (0.25) (0.30)
-0.07
0.09 -0.04
(0.16) (0.20) (0.24)
0.05
0.27
0.34
0.34
0.10
0.10
0.14
Source: Authors’ regressions.
a. Each row reports coefficients from a regression of quarterly revisions to forecasts of the unemployment gap or CPI inflation on a constant and on shocks nt,0 through nt,4, where nt,0 is the monetary policy
shock that occurs contemporaneously with announcement t, and the remaining shocks nt,j are forward
guidance shocks indicating the change in monetary policy announced at t to occur in quarter j. Standard
errors are in parentheses. Asterisks indicate statistical significance at the *10 percent, **5 percent, and
***1 percent level.
b. Each forecast revision is expressed as the forecast value for the period t + j outcome made at time
t + j - n minus the same forecast value made at time t + j - n - 1, where t + j is the subscript and n and
n + 1 are the superscripts.
We find one aspect of the results in table 8 puzzling: the forward guidance shocks have much larger estimated effects on bond yields than does
the contemporaneous monetary policy shock, but the only substantial
difference between nt,j and nt,0 is a j-quarter implementation delay. If the
Treasury rates correspond to the appropriate average of expected shortterm rates plus a term premium, and the forward guidance affects only the
expected short-term rates, then the responses should be nearly identical.
The fact that they are not strongly suggests that our identified forward guidance shocks are affecting term premiums. Fully exploring this intriguing
result lies beyond the scope of the present paper.
Table 9 reports the results from regressing the eight forecast revisions
against a constant and the five n’s. With rational expectations, the constant
36
Brookings Papers on Economic Activity, Spring 2012
term should be irrelevant. It is indeed so for three of the four unemployment forecast revisions, but the Blue Chip forecasters consistently made a
small (but statistically significant) 7-bp error in their final unemployment
forecast. We see similar small but systematic errors in inflation expectations. The slope coefficients’ standard errors are quite large (on the order
of 20 to 30 bp), but nevertheless many of the coefficients on nt,1 in the
unemployment regressions are negative and statistically significant. That
is, promises of more-restrictive policy in the next quarter are associated
with reductions in unemployment expectations. Although the analogous
coefficients from the inflation regressions are not statistically significant, it
is also worth noting that they are positive.
Of course, the New Keynesian model requires that reductions to current and future interest rates be unanticipated if they are to lower expected
unemployment and raise expected inflation, so the negative reaction of
unemployment to nt,1 clearly cannot be interpreted as the direct macroeconomic effect of unanticipated forward guidance. However, neither can it
be interpreted as reflecting simple reverse causality from publicly known
macroeconomic circumstances to monetary policy, because the interest rate
rule accounts for typical monetary policy choices given expectations of
unemployment and inflation. One possibility worth considering is that the
effects arise because the FOMC systematically responds to recent revisions
in expectations.
To understand this further, consider the following augmented interest
rate rule:
(4)
rt = µ + ρ1rt −1 + ρ2 rt − 2 + (1 − ρ1 − ρ2 ) ( φπ π t + φu u t )
M
+ η( u t − u tL ) + ∑ vt − j , j .
j=0
Here h < 0 measures the extent to which the FOMC reacts to unemployment
news received over the last L quarters, specified here as u~ t - u~Lt . One might
suppose that h will be large and negative if the FOMC becomes systematically worried about falling behind the curve following unemployment
surprises. If L ≤ M, then the newly added term in equation 4 is orthogonal
to the instruments we used for estimation, so its presence will not affect our
estimates of r1, r2, fp, and fu. However, it will change the inferred values
of the interest rate rule’s expected intercept, and through this will influence the estimated n’s. Under this interpretation of the results in table 9,
the FOMC’s actions are history dependent. The estimated interest rate rule
campbell, evans, fisher, and justiniano
37
states the typical policy stance given economic conditions forecasted 4
quarters in advance, but the FOMC would respond more aggressively to
the same set of circumstances if it forecasted them only shortly before their
arrival.22
II.F. Summary
What does the analysis of forward guidance identified from a standard
interest rate rule tell us? First, and perhaps most important for the potential viability of forward guidance–based strategies today, the public and
the FOMC together have extensive experience with the communication of
relatively short term forward guidance. Indeed, the FOMC used forward
guidance to signal its acceleration of accommodation in late 2001 and early
2002. Overall, the public anticipated about 40 percent of the variance in the
interest rate rule’s disturbance 3 or 4 quarters in advance. Second, unanticipated accommodative forward guidance reduces the interest rates relevant
for households’ and firms’ economic decisions. That is, it seems possible
for the FOMC to influence longer-term interest rates that are outside of
its direct control by communicating its intention to lower the short-term
policy rate persistently.
III. Using Odyssean Forward Guidance
The foregoing analysis suggests that the FOMC has experience successfully communicating its intended future behavior in response to prevailing
macroeconomic conditions. We interpret this to mean that communication
difficulties do not present an insurmountable barrier to monetary policies
based on Odyssean forward guidance, and that therefore it is worth considering the practical consequences of adopting such policies. Currently,
the FOMC has an extraordinary degree of forward guidance in place with
its “late 2014” statement language. In this section we investigate the consequences of interpreting that language as Odyssean forward guidance
that implements the policy recommendations of Eggertsson and Woodford
(2003) and others. There are legitimate concerns that forward guidance
22. We cannot estimate coefficients like h in equation 4 by regressing the measured
values of nt on expectations revisions, because the true values of nt should be endogenously
correlated with the expectations revision. Since the expectations revision is uncorrelated
over time virtually by construction, neither can we employ an instrumental variables estimator with lagged information as instruments. This leads us to believe that the cross-equation
restrictions of structural models will be essential for identification and estimation of the real
effects of forward guidance.
38
Brookings Papers on Economic Activity, Spring 2012
of this kind places the FOMC’s mandated price stability goal at risk. We
consider these by forecasting the path of the economy under the present
forward guidance and subjecting that forecast to two upside risks: higher
inflation expectations and faster deleveraging by households and firms. We
undertake this analysis using the medium-scale DSGE model developed at
the Federal Reserve Bank of Chicago for just such a purpose.
Evans (2011) has proposed conditioning the FOMC’s forward guidance
on outcomes of unemployment and inflation expectations. Under his proposal, the FOMC would announce specific conditions under which it will
begin lifting its policy rate above zero: either unemployment falling below
7 percent or medium-term expected annual inflation rising above 3 percent would trigger liftoff from the ZLB. Bright-line threshold rules such
as these are designed to maintain low interest rates even as the economy
begins expanding on its own (as prescribed by Eggertsson and Woodford
2003) while providing safeguards against unexpected developments that
might put the FOMC’s price stability goal in jeopardy. We illustrate that
such conditioning, if credible, could be helpful in limiting the inflationary
consequences of an unexpectedly early end to the postcrisis deleveraging.
Our conclusions obviously depend on the assumed structure of the model
economy and the values we assign its parameters. One might therefore
doubt the usefulness of our model-based experiments, since there is little
consensus on what the “right” structural model is, and even when there is
agreement on the model, there is often disagreement over its parameter
values. Nevertheless, we believe our experiments are both interesting and
relevant to policy, for at least two reasons. First, the model is very similar
to other widely used models and is essentially the standard structural tool
for monetary policy analysis in the United States and around the world.
Second, the model’s parameters are estimated using a rich array of macroeconomic data so that our analysis has a firm empirical grounding.
We begin by briefly describing the model, its estimation, and how we
calibrate it to the current policy environment. Then we present our baseline forecast and the consequences for monetary policy of two alternative
scenarios.
III.A. The Model
The model is adapted from Justiniano and others (2011) and thus closely
resembles many other medium-scale empirical New Keynesian models.23 A
single representative household owns all firms and supplies the economy’s
23. The model is described in more detail in Brave and others (2012).
campbell, evans, fisher, and justiniano
39
labor. Final goods are produced with differentiated intermediate goods,
which themselves are produced with capital and differentiated labor. The
intermediate goods market and the labor market are monopolistically competitive. Prices of both kinds of differentiated inputs are sticky and are
subject to partial indexation.24 Hence standard forward-looking Phillips
curves connect wage and price inflation with the marginal rates of substitution between consumption and leisure and marginal cost, respectively.
Other frictions include endogenous capacity utilization, costs of adjusting
investment growth, and internal habit preferences, where “internal habit”
refers to diminishing current utility in lagged own consumption. The combination of all these features is very close to that in models by Lawrence
Christiano, Martin Eichenbaum, and Evans (2005), Frank Smets and Rafael
Wouters (2007), and many others, so that knowledge of these models is
sufficient for understanding the results.
The model has one feature that distinguishes it from other New Keynesian frameworks: the monetary policy interest rate rule.25 This rule is
given by equation 1, except that we set r2 = 0 and replace u~t with the
policy-relevant output gap, ~
yt.26 The policy-relevant measure of inflation
in equation 1 is defined by
(5)
π t =
1 2
∑ E t πˆ t + j − πˆ *t ,
4 j =−1
where “^”denotes deviation from steady state. Equation 5 says that policyrelevant inflation is the deviation of a 4-quarter average of inflation from
the time-varying inflation anchor p̂*. The model’s inflation anchor varies
exogenously and follows an AR(1) process. It is included to account for
low-frequency movements in inflation and to consider policy experiments
in which inflation expectations become “unanchored.” The 4-quarter moving average of inflation includes both lagged, current, and future values of
inflation. The monetary authority uses the structure of the model to forecast
the future terms.
24. In each period wages and prices have a constant probability of being optimally reset;
otherwise they are exogenously indexed to a convex combination of steady-state inflation,
last period’s inflation, and (for wages) productivity growth.
25. The model and estimation involve other unique features, but these do not change the
model’s shock propagation mechanisms, which continue to resemble those in other mediumscale New Keynesian models. The model includes a financial accelerator as in Gilchrist,
Ortiz, and Zakrajšek (2011), but this ends up being unimportant for the results.
26. In future work we intend to consider the case where r2 ≠ 0.
40
Brookings Papers on Economic Activity, Spring 2012
We define the output gap as
y t =
(6)
(7)
{
1 2
∑ E xˆ ,
4 j = −1 t t + j
}
E t 1 + λ (1 − L ) (1 − F ) xˆ t = E t λ (1 − L ) (1 − F ) ŷt
2
2
2
2
where L and F denote lags and leads, respectively, and l is a smoothing
parameter. Equation 6 defines the output gap as a 4-quarter moving average of detrended model output. Following Vasco Cúrdia and others (2011),
the monetary authority detrends output using the filter given by equation 7.
(We consider only stationary solutions.) This detrending approximates
Hodrick-Prescott filtering. The moving average of filtered output has the
same lead-lag structure as inflation and so also includes forward-looking
terms, which embody news about inflation and the output gap up to 2 quarters ahead.
We use the GSS factor structure for the forward guidance shocks
in equation 1. In particular, we allow there to be a target factor and a
path factor driving forward guidance, both of which are independent
and identically distributed over time. All current and forward guidance
shocks load onto the target factor, and all but the contemporaneous policy shock load onto the path factor. Corresponding to each current and
forward guidance shock there is also an additive idiosyncratic shock. For
the precrisis sample we set M = 4 in equation 1 and estimate the factor
loadings, the two factor variances, and variances for the idiosyncratic
shocks at each horizon of forward guidance. Agents in the model therefore see a credible commitment to deviate from the typical response of
policy to current economic conditions going out 4 quarters. Within the
context of the model, the forward guidance shocks are entirely Odyssean
because they are a (credible) commitment to a future action.
We identify the contemporaneous, forward guidance, and inflation
anchor shocks using data on the federal funds rate, federal funds rate futures
prices, and long-run (10-year) inflation expectations taken from the Survey
of Professional Forecasters. The current policy shock moves the current
rate more than it does future rates, whereas the forward guidance and the
inflation anchor shocks move expected future federal funds rates more than
they do the current rate. This difference is a key source of identification.
campbell, evans, fisher, and justiniano
41
Both the inflation anchor and the forward guidance shocks influence inflation, with the effects of the latter arising through the Phillips curve. We
assume that the inflation anchor is very persistent, so that the effects of
forward guidance shocks on inflation expectations are comparatively more
concentrated at shorter horizons. As a result, the forward guidance shocks
are identified from changes in future rates that are larger than changes in
the current rate and are associated with only small movements in long-run
inflation expectations. We do not use the Blue Chip data to identify forward
guidance in the model because we want to consider horizons of forward
guidance beyond 1 year during the period in which the ZLB is binding.
A natural objection to using forward guidance as a tool for generating
additional monetary accommodation is that, by doing so, the monetary
authority risks inflation expectations becoming unhinged. In our sample,
inflation expectations exhibit a downward trend, so we strongly suspect
that episodes of forward guidance raising long-run inflation expectations
are absent from our precrisis sample. That said, one needs to be wary of
this possibility in the current environment.
In addition to the monetary policy shocks, the model’s fluctuations
are driven by eight “structural” shocks. With one exception noted below,
these shocks are assumed to follow an AR(1) process. Four of these shocks
move real GDP and inflation (as measured by the GDP deflator) in the
same direction on impact, so we refer to these as demand shocks. One, the
discount shock, changes households’ rate of time discounting. Another two
are financial disturbances: the spread shock generates fluctuations in the
external finance premium beyond the level warranted by current economic
conditions, and the net worth shock generates exogenous fluctuations in
private balance sheets.27 The fourth demand shock, called the government
shock, is a shock to the sum of government spending, net exports, and
the change in valuation of inventories. Four other shocks move real GDP
and inflation in opposite directions on impact, and so we call these supply shocks. These shocks directly change neutral technology, investmentspecific technology, markups of intermediate goods prices, and households’
disutility from labor. The last of these is assumed to follow an ARMA(1,1)
process, to parsimoniously address low-frequency dynamics in hours
worked and high-frequency variation in hourly wages. Other shocks that
27. These shocks enter because of the financial accelerator mentioned earlier. The net
worth shock plays a negligible role in fluctuations, but the spread shock is a major driver
of fluctuations. The model propagates the spread shock essentially as it does a shock to the
marginal efficiency of investment identified using spread data.
42
Brookings Papers on Economic Activity, Spring 2012
are of small importance in accounting for the data are shocks that do not
affect agents’ decisions: idiosyncratic shocks to the various price measures
used in estimation, and measurement error in the two financial variables
described below.28
III.B. Estimation
We use a two-step procedure to assign values to our DSGE model’s
parameters. First, we estimate the model over the period from 1989Q2
(when federal funds futures contract data begin) to 2007Q2 (just before
the onset of the financial crisis) under the assumption that forward guidance extends out 4 quarters. Second, for the period 2007Q3–2011Q4 we fix
the non–forward guidance parameters at their estimated values (with four
exceptions highlighted below) and reestimate forward guidance under the
assumption that it extends out 10 quarters. Our policy experiments are based
on this new set of monetary policy parameters, but the model’s determination of the state of the economy takes into account the data from before
2007Q4 as well as the parameter values that were in force at that time.
Our estimates for the period 1989Q2–2007Q2 imply that most fluctuations are driven by the demand shocks.29 The data used to estimate the
model include growth rates of nominal GDP per capita, consumption, and
investment; hours per capita worked in the nonfarm business sector; nominal compensation per hour worked in nonfarm business; the GDP deflator; the deflators corresponding to model-based measures of consumption
and investment; the core personal consumption expenditures (PCE) deflator; core CPI; 10-year-ahead forecasts of CPI; an interest rate spread; the
ratio of private credit to GDP; the federal funds rate; and contemporaneous
expectations of the federal funds rate 1 to 4 quarters hence. Consumption
is measured as consumption of nondurable goods and services, and investment includes business fixed investment, residential investment, and PCE
on durable goods.30 The interest rate spread is a weighted average of high-
28. Model-consistent measures of consumption prices do not correspond well with either
of the measures commonly referenced by policymakers and market participants, core PCE
and core CPI. We use a factor structure to model three consumption price series: the two
popular core measures and the measure designed to be consistent with the model. Doing
this delivers predictions for core PCE and core CPI and limits the structural impact of highfrequency fluctuations in inflation that are likely driven by measurement error. Model-based
inflation is identified with the common factor.
29. Technical details of the estimation are discussed in Brave and others (2012).
30. The remaining components of aggregate expenditures—government spending, net
exports, and private inventory accumulation—are modeled as the government shock.
campbell, evans, fisher, and justiniano
43
yield corporate and mortgage-backed bond spreads over 10-year Treasuries
and an asset-backed bond spread over 5-year Treasuries, where the weights
equal the shares of nonfinancial business, household mortgage, and household consumer debt in a measure of total private credit that includes both
households’ and nonfinancial businesses’ debts.
Brave and others (2012) report the parameter estimates in more detail.
Here we highlight two sets of parameters that have important implications for the outcomes of the policy experiments. First, the monetary
policy rule displays a high degree of interest rate smoothing, the inflation
gap coefficient obeys the Taylor principle, and the output gap coefficient
is smaller than the coefficient for inflation. Reflecting the downward
trend in inflation over our sample, the inflation anchor is very persistent.
The plausibility of the policy rule depends in part on the nature of the
output gap in the rule. Brave and others (2012) demonstrate that the model’s
output gap corresponds well to the gap published by the Congressional
Budget Office.
Second, the estimated model has large nominal and real rigidities. Partly
because of the sample over which it is estimated, the slope of the price
Phillips curve is very small, about an order of magnitude smaller than
single-equation estimates (for example, those of Galí and Gertler 1999 and
Eichenbaum and Fisher 2007). The wage Phillips curve slope is also small
but more in line with estimates that do not rely on the full structure of the
model, such as those by Argia Sbordone (2006). Our estimates imply that
there is limited feedback from aggregate activity to wage or price inflation
in the model. The estimated real rigidities as implied by the elasticity of
capacity utilization, investment adjustment costs, and habit are similar in
magnitude to other estimates in the literature (for example, Justiniano and
others 2011) and impart considerable inertia in response to shocks.
III.C. Policy Experiments
The macroeconomic outcomes from 2007Q3 to 2011Q4 are unusual
compared with those of the period used to estimate the model. Therefore, to
conduct policy experiments relevant to the current economic environment,
we calibrate some of the model’s parameters and reestimate the effects of
forward guidance. This reestimation is particularly important because of
the relatively long horizon over which forward guidance has been issued
by the FOMC during the recent period.
We calibrate three parameters for the period 2007Q3–2011Q4: the persistence of the discount shock, the variance of the inflation anchor shock,
and the coefficient on the output gap in the policy rule. To capture the idea
44
Brookings Papers on Economic Activity, Spring 2012
that deleveraging by households and firms following the financial crisis
is unusually slow, we raise the persistence of the discount shock from its
estimated value in the precrisis sample.31 Consequently, the model sees discount shocks playing a larger role since 2007Q2 than at other times, leading to much lower aggregate demand at the end of the sample. Essentially
the model interprets much of the weakness in the data as reflecting agents’
desires to save much more than they have at other times under similar conditions. We set the variance of the inflation anchor innovation to one-fourth
its estimated value from the precrisis period. This choice is motivated by
the fact that inflation expectations exhibit a downward trend in the first part
of our sample but have fluctuated considerably less since then. Finally, we
assume a coefficient on the output gap in the model’s policy rule that is
three times the size of the precrisis estimate. Our motivation here is that the
FOMC’s policy response to a very large recession may be more aggressive
than to a modest one. Together these assumptions increase the likelihood
that the ZLB is binding in any given quarter since 2007Q3.
Given the calibrated parameters and precrisis estimates for the remaining parameters excluding forward guidance and the discount shock’s
variance, we reestimate the factor loadings, factor variances, and idiosyncratic variances that characterize forward guidance as well as the discount
shocks’ variance over the period 2007Q3–2011Q4 under the assumption
that forward guidance extends out 10 quarters.32 Our estimation of forward guidance in this period uses expected future federal funds rates going
out 10 quarters from each date in the sample. With estimates in hand and
data for this period, the Kalman smoother is used to back out the model’s
interpretation of the shocks hitting the economy since the crisis and their
implications for the model’s state variables as of 2011Q4. One important
implication of our calibration and estimated forward guidance is that the
model sees the ZLB as binding from 2008Q4 until the end of our sample
in 2011Q4.33 At this last date the model can be used to generate a forecast
under the assumption that no further shocks hit the economy. This is our
baseline forecast.
31. The discount factor is commonly used to model episodes in which the ZLB is binding. See, for example, Christiano and others (2011).
32. We reestimate the discount shock’s variance to ameliorate concerns that we have
imposed excessive weight on this shock in explaining the crisis.
33. We say the ZLB is binding at any given date if, when all but the forward guidance
factor shocks have been fed into the model to generate a conditional forecast beginning in
2008Q3, the forecasted path of the federal funds rate at each date would be below zero for at
least one period at short horizons.
45
campbell, evans, fisher, and justiniano
Figure 5. Baseline and Alternative Projections
GDP growth
Percent per year
10
Higher expected inflation
8
Shorter deleveraging
Baseline forecast
6
4
2
Federal funds rate
Percent per year
6
5
4
3
Actual
2
1
Average, 1989Q2–2007Q2
2012
2013
0
2014
Core PCE inflation
Percent per year
2012
2013
2014
Log of hours per capita
Log points
0
3
–2
–4
2
–6
1
–10
–8
–12
2012
2013
2014
–14
2012
2013
2014
Source: Authors’ calculations.
Figure 5 displays the baseline forecast along with forecasts corresponding
to two alternative scenarios described below. The horizontal line in each plot
indicates the long-run average of the variable in question over the sample
1989Q2–2007Q2 (the logarithm of hours per capita has a mean that is very close
to zero). The forward guidance in the baseline forecasts has been estimated
to fit the federal funds rate futures path through mid-2014, after which the
model predicts a mild liftoff in the funds rate to about 1 percent at the end of
2014. This path is roughly in line with the “late 2014” forward guidance in
the January and March 2012 FOMC statements. Corresponding to this path
for the funds rate, the baseline forecast calls for growth slightly above trend
for 2012, returning to trend in 2013 and 2014. Growth is sufficiently tepid
that the log of hours per capita is still 10 log points below its steady-state
level by the end of the forecast horizon. Core PCE inflation, after initially
dropping, is forecasted to rise slowly toward its long-run average.
46
Brookings Papers on Economic Activity, Spring 2012
Figure 6. Inflation and Unemployment in the Baseline Forecast
Core PCE inflation (percent per year)
3.0
2.5
Policy objective
2.0
Federal funds rate above ZLB
1.5
2011Q4
1.0
0.5
5.5
6.0
6.5
7.0
7.5
Unemployment rate (percent)
8.0
8.5
Source: Projections of the authors’ model described in the text.
Figure 6 shows the baseline forecast in inflation-unemployment space.34
The horizontal bar represents the FOMC’s policy objective of 2 percent
annual inflation, as described in the FOMC document “Longer-Run Goals
and Policy Strategy,” and the “central tendency” of longer-run unemployment of 5.2 to 6.0 percent reported in the January 2012 release of FOMC
participants’ economic projections. The 2011Q4 launch date for the forecast is labeled, with the economy’s path proceeding from there. The smaller
dots along the path indicate the period of a near-zero federal funds rate,
and the two dots at the far end of the path indicate forecast dates where the
34. Our model does not have unemployment in it. However, an ordinary least squares
regression of unemployment on hours per capita fits extremely well. We use this regression
model to map our forecast for hours per capita into a forecast for unemployment.
campbell, evans, fisher, and justiniano
47
federal funds rate has risen above the ZLB. The bright-line thresholds of
7 percent unemployment and 3 percent inflation are also shown.
In this baseline forecast, core inflation has moved closer to the FOMC’s
explicit objective by the end of 2014. However, unemployment at that date
seems high relative to any rate that would be consistent with the FOMC’s
mandated goal of maximum sustainable employment. Lengthening the
period that the federal funds rate is kept at zero would bring policy closer to
the optimum identified by Eggertsson and Woodford (2003) and Werning
(2012). The FOMC may be disinclined to push the limit of monetary policy
accommodation very far in this dimension, however. Although calendardate communications may have an Odyssean component, most market
analysis seems to interpret the dates as Delphic communications, possibly
limiting their stimulating effect. Finding acceptable bright-line thresholds
might impart a larger commitment to accommodation. Since the forecast
does not breach either the unemployment or the inflation threshold in this
baseline scenario, the threshold rule would prescribe keeping the funds rate
low for a longer period.
It is worth emphasizing that beyond providing additional Odyssean forward guidance, such a threshold rule offers a risk management approach to
guarding against unforeseen circumstances. To illustrate this point, we consider two experiments that simulate the effects of developments that give
rise to greater inflation concerns. In each case we calculate the model’s
forecast from 2011Q4 onward under the assumption that an unanticipated
event occurs in 2012Q1. The state of the economy in 2011Q4 includes all
prior realizations of forward guidance, and agents in the model foresee
exceptionally low interest rates through to late 2014. Our scenarios evaluate the consequences of maintaining this policy regardless of developments
that could lead the FOMC to start raising the federal funds rate earlier. We
do not impose the threshold policy in either scenario. Rather, we simply
monitor the boundaries to examine whether such conditional forward guidance would call for a liftoff from the ZLB sooner than currently anticipated.
For each scenario we assume either a permanent change in a single
model parameter or the realization of a shock for one period. In the scenario
with a parameter change, we resolve the model and use this solution for
the associated forecast. In both scenarios we compute the forecast starting
from the same estimated state of the economy used to construct the baseline forecast. In the scenario with a sudden increase in long-run inflation
expectations, the unanticipated event is an unusually large and persistent
innovation to the inflation anchor. We assume a single innovation to the
inflation anchor that generates an immediate increase in long-run inflation
48
Brookings Papers on Economic Activity, Spring 2012
expectations of 1 percentage point.35 In the rapid deleveraging scenario,
we assume that the persistence of the discount rate shock drops from its
calibrated level of 0.95 to its precrisis level of 0.75 but do not consider any
additional shocks. In this scenario, past realizations of the discount shock
die out much sooner than anticipated in the baseline forecast (the half-life
of a discount shock declines from 3.4 years to 2.4 quarters.)
Each scenario involves solving for the forward guidance that reproduces
the expected funds path through 2014Q2. This is accomplished by setting
one of the idiosyncratic shocks to zero and then solving for the realization of the target and path factors in the first period, plus the other nine
idiosyncratic shocks such that the funds path is matched exactly through
2014Q2. (We apply the estimated factor loadings underlying the baseline
forecast to calculate the forward guidance shocks.) As figure 5 illustrates,
both alternative scenarios generate fast growth immediately: faster deleveraging occurs through a less contractionary discount factor, and higher
expected inflation through lower real interest rates. Therefore, maintaining
the funds rate path requires very large expansionary realizations of the path
factor—essentially large expansionary forward guidance. With this large
amount of monetary accommodation in place, annual inflation rises above
2 percent in both scenarios, although hours per capita remain relatively
low. Presumably less expansionary monetary policy, involving an earlier
liftoff of the funds rate from zero, would be required to forestall this higher
inflation, but this would be at the expense of an even weaker labor market.
Figures 7 and 8 show the two alternative scenarios in inflationunemployment space. These figures are similar to figure 6 except that they
also include the baseline forecast for comparison. Under faster deleveraging, unemployment falls faster and inflation rises by more than in the
baseline. The economy crosses the 7 percent unemployment threshold in
2012Q3 and reaches the 3 percent inflation threshold in late 2013. Therefore, adherence to the 7/3 threshold policy dictates liftoff from the ZLB in
late 2012. Given the improvement in the economy and labor markets, an
earlier exit seems palatable.
We now consider the higher-expected-inflation scenario. Note that generating the increase in inflation expectations in this scenario requires a
shock that is more than 4 standard deviations of the inflation anchor innovation as estimated in the precrisis sample. The resulting forecast, condi35. Given the high persistence of the inflation anchor, the increase in average expected
inflation over the next 40 quarters is actually hump shaped, and therefore higher in later
quarters.
49
campbell, evans, fisher, and justiniano
Figure 7. Inflation and Unemployment in the Forecast with Shorter Deleveraging
Core PCE inflation (percent per year)
Federal funds rate above ZLB
3.0
2.5
Policy objective
2.0
1.5
2011Q4
1.0
Baseline
0.5
5.5
6.0
6.5
7.0
7.5
Unemployment rate (percent)
8.0
8.5
Source: Projections of the authors’ model described in the text.
tioning on exceptionally low rates through at least the next 10 quarters,
does generate a boom in GDP growth. However, because of the strong real
and nominal rigidities we have estimated, neither unemployment nor inflation crosses its threshold within the next 3 years. The unemployment rate
skirts its 7 percent threshold without crossing it, and inflation remains well
below its 3 percent threshold, through the end of 2014. Although the 7/3
threshold policy would dictate keeping rates at the ZLB in this scenario,
the turn in the direction of unemployment toward the end of the forecast
horizon is worrisome.
This scenario illustrates a striking feature of New Keynesian models
estimated using post-1970s data. Because of the very flat price Phillips
curve, very large innovations to inflation expectations do not lead to high
inflation even with extraordinarily accommodative monetary policy, at
least over a 3-year horizon. This result depends on the assumed credibility
50
Brookings Papers on Economic Activity, Spring 2012
Figure 8. Inflation and Unemployment in the Forecast with Higher Expected Inflation
Core PCE inflation (percent per year)
3.0
Federal funds rate above ZLB
2.5
Policy objective
2.0
1.5
2011Q4
1.0
Baseline
0.5
5.5
6.0
6.5
7.0
7.5
Unemployment rate (percent)
8.0
8.5
Source: Projections of the authors’ model described in the text.
of the model’s policy rule and invariance of price setting behavior to inflation expectations. If attempted use of Odyssean forward guidance weakens
credibility or changes price setting behavior, this kind of policy experiment
might be very misleading. Nevertheless, nothing in the experience of the
last 25 years suggests that a persistent change in inflation expectations necessarily generates a destabilizing loss of credibility.
IV. Conclusion
The empirical context we have provided shows that the FOMC has extensive experience at broadcasting its intended responses to macroeconomic
developments. Indeed, macroeconomic forecasters and market participants anticipate about 80 percent of the FOMC’s deviations from a simple
campbell, evans, fisher, and justiniano
51
interest rate rule. These communications have not been limited to a single
“tight-versus-loose” dimension. The FOMC successfully informed markets
that it would accelerate its accommodation in late 2001 and early 2002 and
accelerate its removal. Our results also show that surprises associated with
FOMC policy announcements substantially influence Treasury bond rates,
corporate borrowing rates, and private macroeconomic forecasts. News
of substantial monetary tightening raises interest rates as expected, but it
also raises inflation forecasts and lowers unemployment forecasts. This
counterintuitive finding suggests to us that private forecasters believe that
nonpublic information held by the Federal Reserve about future economic
conditions instigates some FOMC actions that were unanticipated by the
public. That is, the public sometimes imputes Delphic content to policy
announcements that are not explicitly tied to economic fundamentals.
As expressed in its April 2012 statement, the most recent as of this writing, the FOMC “. . . anticipates that economic conditions—including low
rates of resource utilization and a subdued outlook for inflation over the
medium run—are likely to warrant exceptionally low levels for the federal funds rate at least through late 2014.” We began this paper by asking
whether this statement reflects an Odyssean commitment to lower rates or
a Delphic forecast of economic conditions and the FOMC’s likely response
to them. Our empirical results reassure us that communications difficulties
present no insurmountable obstacle to the FOMC stressing the Odyssean
interpretation and thereby providing additional monetary accommodation,
but other objections to such a policy remain. In particular, one might worry
that an Odyssean commitment to low rates places the FOMC’s price stability mandate in jeopardy.
We have addressed this concern by using the Chicago Federal Reserve
Bank’s estimated DSGE model to simulate two adverse scenarios. In the
first, the deleveraging process presently keeping the economy at the ZLB
accelerates and finishes sooner than expected, and in the second, long-run
inflation expectations suddenly rise 1 full percentage point. We compare
both simulations with the “bright-line” threshold policy proposal of Evans
(2011), which calls for rate increases to begin when either unemployment
falls below 7 percent or medium-term expected annual inflation rises above
3 percent. With faster deleveraging beginning in 2012Q1, the unemployment rate falls below its threshold for triggering rate increases in 2012Q3.
In this case the policy provides useful insurance against the inflationary
consequences of an unforeseen economic recovery. With an exogenous rise
in inflation expectations occurring in 2012Q1, the economy comes close
to (but does not cross) the unemployment threshold at the start of 2014
52
Brookings Papers on Economic Activity, Spring 2012
and comes nowhere near the inflation threshold. We conclude from these
experiments that the risks of Odyssean forward guidance to the Federal
Reserve’s price stability mandate can be managed with such conditional
forward guidance.
ACKNOWLEDGMENTS We are grateful to Marco Bassetto and Spencer
Krane for many stimulating discussions on forward guidance, and to Charles
Calomiris, Hesna Genay, David Reifschneider, John Williams, Michael
Woodford, and the editors for comments on earlier drafts. We thank Refet
Gürkaynak and Eric Swanson for sharing their data with us. Jacob Fabina,
Matt Olson, and Christine Ostrowski provided expert research assistance.
The views expressed here are those of the authors and do not necessarily
represent those of the Federal Reserve Bank of Chicago, the Federal Reserve
System, or its Board of Governors.
campbell, evans, fisher, and justiniano
53
References
Brave, Scott, Jeffrey Campbell, Jonas Fisher, and Alejandro Justiniano. 2012. “The
Chicago Fed DSGE Model.” Working Paper no. 2012–02. Federal Reserve
Bank of Chicago.
Christiano, Lawrence J., Martin Eichenbaum, and Charles L. Evans. 2005. “Nominal Rigidities and the Dynamic Effects of a Shock to Monetary Policy.” Journal
of Political Economy 113, no. 1: 1–45.
Christiano, Lawrence, Martin Eichenbaum, and Sergio Rebelo. 2011. “When Is the
Government Spending Multiplier Large?” Journal of Political Economy 119,
no. 1: 78–121.
Cúrdia, Vasco, Andrea Ferrero, Ging Cee Ng, and Andrea Tambalotti. 2011. “Evaluating Interest Rate Rules in an Estimated DSGE Model.” Working Paper no. 510.
Federal Reserve Bank of New York.
Eggertsson, Gauti B., and Michael Woodford. 2003. “The Zero Bound on Interest
Rates and Optimal Monetary Policy.” BPEA, no. 1: 139–211.
Eichenbaum, Martin, and Jonas D. M. Fisher. 2007. “Estimating the Frequency of
Price Re-optimization in Calvo-Style Models.” Journal of Monetary Economics
54, no. 7: 2032–47.
Ellingsen, Tore, and Ulf Söderström. 2001. “Monetary Policy and Market Interest
Rates.” American Economic Review 91, no. 5: 1594–1607.
Evans, Charles. 2011. “The Fed’s Dual Mandate Responsibilities and Challenges
Facing U.S. Monetary Policy.” Remarks delivered at the European Economics
and Financial Centre Distinguished Speaker Seminar, London, September 7.
———. 2012. “Monetary Policy in a Low-Inflation Environment: Developing a
State-Contingent Price-Level Target.” Journal of Money, Credit and Banking
44 (Issue Supplement s1).
Gagnon, Joseph, Matthew Raskin, Julie Remache, and Brian Sack. 2010. “LargeScale Asset Purchases by the Federal Reserve: Did They Work?” Working
paper. Federal Reserve Bank of New York.
Galí, Jordi, and Mark Gertler. 1999. “Inflation Dynamics: A Structural Econometric Analysis.” Journal of Monetary Economics 44, no. 2: 195–222.
Gilchrist, Simon, Alberto Ortiz, and Egon Zakrajšek. 2011. “Credit Risk and the
Macroeconomy: Evidence from an Estimated DSGE Model.” Working paper.
Boston University.
Gürkaynak, Refet, Brian Sack, and Eric Swanson. 2005. “Do Actions Speak
Louder Than Words? The Response of Asset Prices to Monetary Policy Actions
and Statements.” International Journal of Central Banking 1, no. 1: 55–93.
Hamilton, James D., Seth Pruitt, and Scott Borger. 2011. “Estimating the MarketPerceived Monetary Policy Rule.” American Economic Journal: Macro
economics 3, no. 3: 1–28.
Justiniano, Alejandro, Giorgio E. Primiceri, and Andrea Tambalotti. 2011. “Investment Shocks and the Relative Price of Investment.” Review of Economic
Dynamics 14, no. 1: 101–21.
54
Brookings Papers on Economic Activity, Spring 2012
Krane, Spencer D. 2011. “Professional Forecasters’ Views of Permanent and Transitory Shocks to GDP.” American Economic Journal: Macroeconomics 3, no. 1:
184–211.
Krishnamurthy, Arvind, and Annette Vissing-Jorgensen. 2011. “The Effects of
Quantitative Easing on Interest Rates: Channels and Implications for Policy.”
BPEA, no. 2: 215–66.
Krugman, Paul R. 1999. “It’s Baaack: Japan’s Slump and the Return of the Liquidity Trap.” BPEA, no. 2: 137–87.
Kuttner, Kenneth N. 2001. “Monetary Policy Surprises and Interest Rates: Evidence from the Fed Funds Futures Market.” Journal of Monetary Economics
47, no. 3: 523–44.
Kydland, Finn E., and Edward C. Prescott. 1977. “Rules Rather Than Discretion:
The Inconsistency of Optimal Plans.” Journal of Political Economy 85, no. 3:
473–92.
Laséen, Stefan, and Lars E. Svensson. 2011. “Anticipated Alternative Policy-Rate
Paths in Policy Simulations.” International Journal of Central Banking 7, no. 3:
1–35.
Reifschneider, David, and John C. Williams. 2000. “Three Lessons for Monetary
Policy in a Low-Inflation Era.” Journal of Money, Credit, and Banking 32, no. 4:
936–66.
Romer, Christina, and David Romer. 2000. “Federal Reserve Information and the
Behavior of Interest Rates.” American Economic Review 90, no. 3: 429–57.
Rudebusch, Glenn D., and John C. Williams. 2008. “Revealing the Secrets of
the Temple: The Value of Publishing Central Bank Interest Rate Projections.”
In Asset Prices and Monetary Policy, edited by J. Y. Campbell. University of
Chicago Press.
Sbordone, Argia. 2006. “U.S. Wage and Price Dynamics: A Limited Information
Approach.” International Journal of Central Banking 2, no. 3: 155–91.
Smets, Frank, and Rafael Wouters. 2007. “Shocks and Frictions in US Business
Cycles: A Bayesian DSGE Approach.” American Economic Review 97, no. 3:
586–606.
Swanson, Eric T., and John C. Williams. 2012. “Measuring the Effect of the Zero
Lower Bound on Medium- and Longer-Term Interest Rates.” Working Paper
no. 2012–02. Federal Reserve Bank of San Francisco.
Taylor, John B. 1993. “Discretion versus Policy Rules in Practice.” CarnegieRochester Conference Series on Public Policy 39: 195–214.
———. 1999. “An Historical Analysis of Monetary Policy Rules.” In Monetary
Policy Rules, edited by J. B. Taylor. University of Chicago Press.
Werning, Iván. 2012. “Managing a Liquidity Trap: Monetary and Fiscal Policy.”
Working paper. Massachusetts Institute of Technology.
Wright, Jonathan. 2012. “What Does Monetary Policy Do to Long-Term Interest
Rates at the ZLB?” Working paper. Johns Hopkins University.
Working Paper Series
A series of research studies on regional economic issues relating to the Seventh Federal
Reserve District, and on financial and economic topics.
Why Has Home Ownership Fallen Among the Young?
Jonas D.M. Fisher and Martin Gervais
WP-09-01
Why do the Elderly Save? The Role of Medical Expenses
Mariacristina De Nardi, Eric French, and John Bailey Jones
WP-09-02
Using Stock Returns to Identify Government Spending Shocks
Jonas D.M. Fisher and Ryan Peters
WP-09-03
Stochastic Volatility
Torben G. Andersen and Luca Benzoni
WP-09-04
The Effect of Disability Insurance Receipt on Labor Supply
Eric French and Jae Song
WP-09-05
CEO Overconfidence and Dividend Policy
Sanjay Deshmukh, Anand M. Goel, and Keith M. Howe
WP-09-06
Do Financial Counseling Mandates Improve Mortgage Choice and Performance?
Evidence from a Legislative Experiment
Sumit Agarwal,Gene Amromin, Itzhak Ben-David, Souphala Chomsisengphet,
and Douglas D. Evanoff
WP-09-07
Perverse Incentives at the Banks? Evidence from a Natural Experiment
Sumit Agarwal and Faye H. Wang
WP-09-08
Pay for Percentile
Gadi Barlevy and Derek Neal
WP-09-09
The Life and Times of Nicolas Dutot
François R. Velde
WP-09-10
Regulating Two-Sided Markets: An Empirical Investigation
Santiago Carbó Valverde, Sujit Chakravorti, and Francisco Rodriguez Fernandez
WP-09-11
The Case of the Undying Debt
François R. Velde
WP-09-12
Paying for Performance: The Education Impacts of a Community College Scholarship
Program for Low-income Adults
Lisa Barrow, Lashawn Richburg-Hayes, Cecilia Elena Rouse, and Thomas Brock
Establishments Dynamics, Vacancies and Unemployment: A Neoclassical Synthesis
Marcelo Veracierto
WP-09-13
WP-09-14
1
Working Paper Series (continued)
The Price of Gasoline and the Demand for Fuel Economy:
Evidence from Monthly New Vehicles Sales Data
Thomas Klier and Joshua Linn
WP-09-15
Estimation of a Transformation Model with Truncation,
Interval Observation and Time-Varying Covariates
Bo E. Honoré and Luojia Hu
WP-09-16
Self-Enforcing Trade Agreements: Evidence from Time-Varying Trade Policy
Chad P. Bown and Meredith A. Crowley
WP-09-17
Too much right can make a wrong: Setting the stage for the financial crisis
Richard J. Rosen
WP-09-18
Can Structural Small Open Economy Models Account
for the Influence of Foreign Disturbances?
Alejandro Justiniano and Bruce Preston
WP-09-19
Liquidity Constraints of the Middle Class
Jeffrey R. Campbell and Zvi Hercowitz
WP-09-20
Monetary Policy and Uncertainty in an Empirical Small Open Economy Model
Alejandro Justiniano and Bruce Preston
WP-09-21
Firm boundaries and buyer-supplier match in market transaction:
IT system procurement of U.S. credit unions
Yukako Ono and Junichi Suzuki
Health and the Savings of Insured Versus Uninsured, Working-Age Households in the U.S.
Maude Toussaint-Comeau and Jonathan Hartley
WP-09-22
WP-09-23
The Economics of “Radiator Springs:” Industry Dynamics, Sunk Costs, and
Spatial Demand Shifts
Jeffrey R. Campbell and Thomas N. Hubbard
WP-09-24
On the Relationship between Mobility, Population Growth, and
Capital Spending in the United States
Marco Bassetto and Leslie McGranahan
WP-09-25
The Impact of Rosenwald Schools on Black Achievement
Daniel Aaronson and Bhashkar Mazumder
WP-09-26
Comment on “Letting Different Views about Business Cycles Compete”
Jonas D.M. Fisher
WP-10-01
Macroeconomic Implications of Agglomeration
Morris A. Davis, Jonas D.M. Fisher and Toni M. Whited
WP-10-02
Accounting for non-annuitization
Svetlana Pashchenko
WP-10-03
2
Working Paper Series (continued)
Robustness and Macroeconomic Policy
Gadi Barlevy
WP-10-04
Benefits of Relationship Banking: Evidence from Consumer Credit Markets
Sumit Agarwal, Souphala Chomsisengphet, Chunlin Liu, and Nicholas S. Souleles
WP-10-05
The Effect of Sales Tax Holidays on Household Consumption Patterns
Nathan Marwell and Leslie McGranahan
WP-10-06
Gathering Insights on the Forest from the Trees: A New Metric for Financial Conditions
Scott Brave and R. Andrew Butters
WP-10-07
Identification of Models of the Labor Market
Eric French and Christopher Taber
WP-10-08
Public Pensions and Labor Supply Over the Life Cycle
Eric French and John Jones
WP-10-09
Explaining Asset Pricing Puzzles Associated with the 1987 Market Crash
Luca Benzoni, Pierre Collin-Dufresne, and Robert S. Goldstein
WP-10-10
Prenatal Sex Selection and Girls’ Well‐Being: Evidence from India
Luojia Hu and Analía Schlosser
WP-10-11
Mortgage Choices and Housing Speculation
Gadi Barlevy and Jonas D.M. Fisher
WP-10-12
Did Adhering to the Gold Standard Reduce the Cost of Capital?
Ron Alquist and Benjamin Chabot
WP-10-13
Introduction to the Macroeconomic Dynamics:
Special issues on money, credit, and liquidity
Ed Nosal, Christopher Waller, and Randall Wright
WP-10-14
Summer Workshop on Money, Banking, Payments and Finance: An Overview
Ed Nosal and Randall Wright
WP-10-15
Cognitive Abilities and Household Financial Decision Making
Sumit Agarwal and Bhashkar Mazumder
WP-10-16
Complex Mortgages
Gene Amromin, Jennifer Huang, Clemens Sialm, and Edward Zhong
WP-10-17
The Role of Housing in Labor Reallocation
Morris Davis, Jonas Fisher, and Marcelo Veracierto
WP-10-18
Why Do Banks Reward their Customers to Use their Credit Cards?
Sumit Agarwal, Sujit Chakravorti, and Anna Lunn
WP-10-19
3
Working Paper Series (continued)
The impact of the originate-to-distribute model on banks
before and during the financial crisis
Richard J. Rosen
WP-10-20
Simple Markov-Perfect Industry Dynamics
Jaap H. Abbring, Jeffrey R. Campbell, and Nan Yang
WP-10-21
Commodity Money with Frequent Search
Ezra Oberfield and Nicholas Trachter
WP-10-22
Corporate Average Fuel Economy Standards and the Market for New Vehicles
Thomas Klier and Joshua Linn
WP-11-01
The Role of Securitization in Mortgage Renegotiation
Sumit Agarwal, Gene Amromin, Itzhak Ben-David, Souphala Chomsisengphet,
and Douglas D. Evanoff
WP-11-02
Market-Based Loss Mitigation Practices for Troubled Mortgages
Following the Financial Crisis
Sumit Agarwal, Gene Amromin, Itzhak Ben-David, Souphala Chomsisengphet,
and Douglas D. Evanoff
WP-11-03
Federal Reserve Policies and Financial Market Conditions During the Crisis
Scott A. Brave and Hesna Genay
WP-11-04
The Financial Labor Supply Accelerator
Jeffrey R. Campbell and Zvi Hercowitz
WP-11-05
Survival and long-run dynamics with heterogeneous beliefs under recursive preferences
Jaroslav Borovička
WP-11-06
A Leverage-based Model of Speculative Bubbles (Revised)
Gadi Barlevy
WP-11-07
Estimation of Panel Data Regression Models with Two-Sided Censoring or Truncation
Sule Alan, Bo E. Honoré, Luojia Hu, and Søren Leth–Petersen
WP-11-08
Fertility Transitions Along the Extensive and Intensive Margins
Daniel Aaronson, Fabian Lange, and Bhashkar Mazumder
WP-11-09
Black-White Differences in Intergenerational Economic Mobility in the US
Bhashkar Mazumder
WP-11-10
Can Standard Preferences Explain the Prices of Out-of-the-Money S&P 500 Put Options?
Luca Benzoni, Pierre Collin-Dufresne, and Robert S. Goldstein
WP-11-11
Business Networks, Production Chains, and Productivity:
A Theory of Input-Output Architecture
Ezra Oberfield
Equilibrium Bank Runs Revisited
Ed Nosal
WP-11-12
WP-11-13
4
Working Paper Series (continued)
Are Covered Bonds a Substitute for Mortgage-Backed Securities?
Santiago Carbó-Valverde, Richard J. Rosen, and Francisco Rodríguez-Fernández
WP-11-14
The Cost of Banking Panics in an Age before “Too Big to Fail”
Benjamin Chabot
WP-11-15
Import Protection, Business Cycles, and Exchange Rates:
Evidence from the Great Recession
Chad P. Bown and Meredith A. Crowley
WP-11-16
Examining Macroeconomic Models through the Lens of Asset Pricing
Jaroslav Borovička and Lars Peter Hansen
WP-12-01
The Chicago Fed DSGE Model
Scott A. Brave, Jeffrey R. Campbell, Jonas D.M. Fisher, and Alejandro Justiniano
WP-12-02
Macroeconomic Effects of Federal Reserve Forward Guidance
Jeffrey R. Campbell, Charles L. Evans, Jonas D.M. Fisher, and Alejandro Justiniano
WP-12-03
5
@FedFRASER