Upload
dinhduong
View
217
Download
0
Embed Size (px)
Citation preview
Discussion ofInflation Expectations - a Policy Tool
by Coibion, Gorodnichenko, Kumar and Pedemonte
Ricardo ReisLSE
ECB Forum, Sintra19th of June, 2018
Survey inflation expectations1. Are consistent with widespread inattention
Survey inflation expectations1. Are consistent with widespread inattention• Many people are clueless, little to no attention.• Disagreement about present and future.• Slow diffusion of information, past experiences linger.• Update in response to news, faster in volatile environment.• Available data (shopping, media) affects expectations.• Last 20 years, many unaware of what ECB does, its targets.
Survey inflation expectations1. Are consistent with widespread inattention• Many people are clueless, little to no attention.• Disagreement about present and future.• Slow diffusion of information, past experiences linger.• Update in response to news, faster in volatile environment.• Available data (shopping, media) affects expectations.• Last 30 years, many unaware of what ECB does, its targets.• Not just limited information, lack credibility, backward looking
Survey inflation expectations1. Are consistent with widespread inattention2. Can be properly measured, with effort
Survey inflation expectations1. Are consistent with widespread inattention2. Can be properly measured, with effort• Have done it for households for decades, firms only recently.• Use distributions for uncertainty (never disagreement).• Wording of “inflation” not relevant.• Design of questions matters, ranges matter, priming.• Sampling on characteristics matters.• Expectations of individual versus aggregate variables matters.
Survey inflation expectations1. Are consistent with widespread inattention2. Can be properly measured, with effort• Have done it for households for decades, firms only recently.• Use distributions for uncertainty (never disagreement).• Wording of “inflation” not relevant.• Design of questions matters, ranges matter, priming.• Sampling on characteristics matters.• Expectations of individual versus aggregate variables matters.• Require ECB involvement and effort, but worthwhile
Survey inflation expectations1. Are consistent with widespread inattention2. Can be properly measured, with effort3. Are correlated with actual economic decisions
Survey inflation expectations1. Are consistent with widespread inattention2. Can be properly measured, with effort3. Are correlated with actual economic decisions• Higher expected spending, higher willingness to spend.• Sometimes, raise prices/wages.• Sometimes, lower hiring/investment.
Survey inflation expectations1. Are consistent with widespread inattention2. Can be properly measured, with effort3. Are correlated with actual economic decisions• Higher expected spending, higher willingness to spend.• Sometimes, raise prices/wages.• Sometimes, lower hiring/investment.• Are not just noise, can be very informative
Survey inflation expectations4. Policy announcements have little effect on them
Survey inflation expectations4. Policy announcements have little effect on them• Given inattention, at best only some people.• Given sticky information, lower frequency.• Given correlation speeches and actions, must disentangle.• Given reverse causality, hard to identify.
Authors’ event study
Inflation Expectations as a Policy Tool? 32
To assess how such inattention to what should be large and visible economic announcements can occur, we consider responses to the following question in the MSC: “During the last few months, have you heard of any favorable or unfavorable changes in business conditions?” We use this question to evaluate how consumers are receiving information about different types of policies. Answers are separated by the type of news. We focus on monetary news to see if announcements are reaching households. To quantify the exposure of these announcements, we use a measure of how the media covered these events. This measure is constructed by counting all the news articles that have the phrase “Federal Reserve” in the New York Times (“Fed news”). We have monthly data for both measures. Figure 9 plots time series of monetary news and Fed news for a 13-month window around the announcements. We can see that these big announcements seem to have been covered by the media (or at least the New York Times), as we see a reaction of the amount of news related to the Federal Reserve. Despite this upsurge of news reports, we see little reaction in terms of households reporting receiving more information about monetary policy. The percentage of households who heard about monetary news changes little and in
Figure 7 Reaction of financial markets to Fed announcements
Notes: These figures show the TED spread (black, thick line) and the 5-year forward inflation rate expectation (red, thin line) at a daily frequency. Source: FRED. Panel A shows the 50-basis-point cut in the policy rate on August 17, 2007. Panel B shows the announcement of the first quantitative easing policy on November 25, 2008. Panel C shows the announcement of the second quantitative easing policy on November 3, 2010. Finally, Panel D shows the announcement of the 2% inflation target by the Federal Reserve on January 25, 2012.
2.4
2.45
2.5
5y e
xpec
ted
infla
tion
.51
1.5
22.
5TE
D s
prea
d
18jul2007 07aug2007 27aug2007 16sep2007
Panel A: August 2007: Cut 50bp
.51
1.5
22.
53
5y e
xpec
ted
infla
tion
1.5
22.
53
TED
spr
ead
26oct2008 15nov2008 05dec2008 25dec2008
TED5y Eπ
Panel B: November 2008: QE1
2.3
2.4
2.5
2.6
2.7
2.8
5y e
xpec
ted
infla
tion
.12
.13
.14
.15
.16
.17
TED
spr
ead
04oct2010 24oct2010 13nov2010 03dec2010
Panel C: November 2010: QE2
2.2
2.3
2.4
2.5
2.6
5y e
xpec
ted
infla
tion
.4.4
5.5
.55
.6TE
D s
prea
d
26dec2011 15jan2012 04feb2012 24feb2012
Panel D: January 2012: 2% IT
Inflation Expectations as a Policy Tool? 31
Figure 6 Change in forecasts in Bloomberg’s SPF
Notes: These figures show the number of changes in predictions made by professional forecasters in a given month according to the survey of professional forecasters conducted by Bloomberg. The vertical lines show relevant events or announcements related to the Federal Reserve. Panel A shows the 50-basis-point cut in the policy rate on August 17, 2007. Panel B shows the announcement of the first quantitative easing policy on November 25, 2008. Panel C shows the announcement of the second quantitative easing policy on November 3, 2010. Finally, Panel D shows the announcement of the 2% inflation target by the Federal Reserve on January 25, 2012.
010
2030
40ch
ange
in p
redi
ctio
n
2006m7 2007m1 2007m7 2008m1 2008m7
Panel A: August 2007: Cut 50bp
010
2030
40ch
ange
in p
redi
ctio
n
2007m10 2008m4 2008m10 2009m4 2009m10
Panel B: November 2008: QE1
010
2030
40ch
ange
in p
redi
ctio
n
2009m10 2010m4 2010m10 2011m4 2011m10
Panel C: November 2010: QE2
010
2030
40ch
ange
in p
redi
ctio
n
2011m1 2011m7 2012m1 2012m7 2013m1
Panel D: January 2012: 2% IT
Authors’ event study
Inflation Expectations as a Policy Tool? 33
some cases we even see declines around the main event. Jointly, this indicates that the increased news coverage in major news media sources is either not seen by most households or ignored by them when they read the news.
Figure 8 Inflation expectations in MSC
Notes: These figures plot the weighted average for the inflation expectation of consumers in the Michigan Survey of Consumers. Panel A shows the 50-basis-point cut in the policy rate on August 17, 2007. Panel B shows the announcement of the first quantitative easing policy on November 25, 2008. Panel C shows the announcement of the second quantitative easing policy on November 3, 2010. Finally, Panel D shows the announcement of the 2% inflation target by the Federal Reserve on January 25, 2012.
12
34
56
7pe
rcen
t
2006m7 2007m1 2007m7 2008m1 2008m7
Panel A: August 2007: Cut 50bp
12
34
56
7pe
rcen
t
2007m10 2008m4 2008m10 2009m4 2009m10
Panel B: November 2008: QE11
23
45
67
perc
ent
2009m10 2010m4 2010m10 2011m4 2011m10
Panel C: November 2010: QE2
12
34
56
7pe
rcen
t
2011m1 2011m7 2012m1 2012m7 2013m1
Panel D: January 2012: 2% IT
Same frequency
May2011 Sep2011 Jan2012 May2012 Sep2012 Jan2013
2.5
5.0
1.5
2.0
3.0
3.5
4.0
4.5
5-YearBreakevenInflationRate
UniversityofMichigan:InflationExpectation
Pe
rce
nt
ShadedareasindicateU.S.recessionsSource:FederalReserveBankofSt.Louis,UniversityofMichigan myf.red/g/kcrS
Survey inflation expectations4. Policy announcements have little effect on them• Given inattention, at best only some people.• Given sticky information, lower frequency.• Given correlation speeches and actions, must disentangle.• Given reverse causality, hard to identify.• Very hard to know, maybe, I just don’t know at high frequencies
Volcker disinflation
Source: Mankiw, Reis, Wolfers, 2004
Inflation targeting and professionals
Figure 1. Box plots of Coefficients of Variation 1/
(1989:10 - 2006:11) 2/
Non targeter Targeter
0
50
100
150
200
Coe
ffic
ient
of V
aria
tion
(Fol
low
ing
Yea
r)
1/ Coefficients of variation were calculated as the interquartile range across forecasters divided by the absolute value of the median using monthly forecasts for next year inflation. The box-plot for non-targeters includes data from: Argentina, France, Germany, Indonesia, Italy, Malaysia, the Netherlands, the United States and Venezuela, and data from targeting countries before they implemented inflation targeting. The box-plot for targeters includes data from: Australia, Brazil, Canada, Chile, Colombia, Mexico, Norway, Peru, South Korea, Spain from 1995 to 1999, Sweden, Switzerland, Thailand and United Kingdom, after they implemented inflation targeting.. 2/ Although the maximum value for the coefficient of variation encompassed by the whiskers is 250, there are 5 observations for the non-targeters not included in the figure, with a maximum value of 800. Source: Data from Consensus Forecasts.
23
Figure 1. Box plots of Coefficients of Variation 1/
(1989:10 - 2006:11) 2/
Non targeter Targeter
0
50
100
150
200
Coe
ffic
ient
of V
aria
tion
(Fol
low
ing
Yea
r)
1/ Coefficients of variation were calculated as the interquartile range across forecasters divided by the absolute value of the median using monthly forecasts for next year inflation. The box-plot for non-targeters includes data from: Argentina, France, Germany, Indonesia, Italy, Malaysia, the Netherlands, the United States and Venezuela, and data from targeting countries before they implemented inflation targeting. The box-plot for targeters includes data from: Australia, Brazil, Canada, Chile, Colombia, Mexico, Norway, Peru, South Korea, Spain from 1995 to 1999, Sweden, Switzerland, Thailand and United Kingdom, after they implemented inflation targeting.. 2/ Although the maximum value for the coefficient of variation encompassed by the whiskers is 250, there are 5 observations for the non-targeters not included in the figure, with a maximum value of 800. Source: Data from Consensus Forecasts.
23
Source: Capistran, Ramos-Francia, 2009
Inflation targeting and professionals
date, fij0 .8 Since the ‘‘treatment’’ dummy DT is assumed exogenous (conditional on the matching of a control group of
observations with the treatment observations), suitable instruments for DT Vij0 are fDT Vij
0;g ;DT fij0 g.
5. Results
As a first pass, this section shows some simple graphical results that illustrate the paper’s main finding. It then goes onto outline the baseline regression results and to summarize a number of robustness checks.
5.1. Graphical results
Fig. 1 plots the change in the mean absolute inflation forecast error (DVij) against the average prior error (Vij0 ) for
observations from IT adoption episodes (the ‘‘treated’’) and from the full pool of ‘‘control’’ episodes.9 The sample of controlsincludes, for each of the 11 IT adoption episodes, all available forecasters in non-IT adoption countries for the period inwhich IT was adopted (unlike in the matching exercise below, where each treated observation is matched with a specificcontrol). The relationship appears to be negative for all forecasters, due to mean reversion. However, there is also strongevidence of an additional negative effect (conditional on the initial forecast error) due to IT adoption, as predicted by themodel (the p-value for the test that the slopes are the same is 0.000).
5.2. Matching results
The results of estimating Eq. (9) are presented in Table 2. Column I presents results with the interaction effectsuppressed, measuring only the levels (or unconditional) effect of IT adoption on forecast errors. Column II presentsestimates for the full equation using OLS; column III presents results using the IV strategy (two-stage least squares, 2SLS).10
Since advanced as well as emerging market economies are included in the sample a dummy for advanced countries isincluded in each specification.11
ARTICLE IN PRESS
-10
-5
0
5
10
chan
ge in
fore
cast
erro
r
0 2 4 6 8prior forecast error
IT adopters Non-adopters
Fig. 1. Change in forecast performance around IT adoption. Change in average absolute forecast errors between the 12 months prior to IT adoption andthe subsequent 12 months plotted against the average absolute forecast error in the 12 months prior to IT adoption. Sample includes 196 forecasters in ITadoption countries and 2,048 forecasters in non-IT adoption countries. Sample imposes a common support in terms of prior forecast error between the ITand non-IT adoption group. Sample is trimmed to exclude outliers (defined as an absolute change in the forecast error in excess of 10 percentage points).
8 Using expected rather than actual inflation eliminates the impact of unexpected inflationary shocks that are likely to be correlated with thetransient component of forecast errors. Since shocks to forecast accuracy should, at least in theory, lead forecasters to over- and underestimate inflationwith equal probability, fij
0 should not be correlated with eij0.
9 A common support for V0 is imposed for IT and non-IT adoption episodes. In addition, outliers (defined as those with an absolute change in theinflation forecast error jDV j in excess of 10 percent) are also dropped. The sample includes 196 observations from IT adoption episodes and 2,048observations from control episodes. The results are similar if the sample is not truncated to exclude outliers.
10 Bertrand et al. (2004) show that clustering the residuals (allowing arbitrary patterns of correlations within groups) is a simple and effective meansof avoiding the problem of overstating the significance level attached to estimated treatment effects that tends to occur when difference-in-differenceestimators ignore within-group correlation. Since the observations from the same ‘‘group’’ (defined over episode and country) are based on forecasts ofthe same variable, one would expect significant within-group correlation in the current study. All results therefore report clustered standard errors, with49 clusters in the baseline sample. Non-clustered standard error estimates (either assuming iid errors or robust estimates) are much lower, as one wouldexpect.
11 For the purposes of this paper, the ‘‘advanced’’ countries in the dataset are Australia, Canada, France, Germany, Italy, Japan, the Netherlands,Norway, Spain, Switzerland, the United Kingdom, and the United States.
C. Crowe / Journal of Monetary Economics 57 (2010) 226–232230
Source: Crowe, 2010
Survey inflation expectations4. Policy announcements have little effect on them5. Better communication can break inattention
Survey inflation expectations4. Policy announcements have little effect on them5. Better communication can break inattention• Revealing past or target inflation changes expectations. Target message to scenario• FOMC statements add little, news adds little. Simple messages do better• Information revealed affects expectations for at best 6 months. Repeat the message• Take the message direct to the target evidence
Survey inflation expectations4. Policy announcements have little effect on them5. Better communication can break inattention• Revealing past or target inflation changes expectations. Target message to scenario• FOMC statements add little, news adds little. Simple messages do better• Information revealed affects expectations for at best 6 months. Repeat the message• Take the message direct to the target evidence• Sensible, but no evidence right now.
Survey inflation expectations1. Are consistent with widespread inattention Yes2. Can be properly measured, with effort Yes3. Are correlated with actual economic decisions Yes4. Policy announcements have little effect on them No5. Better communication can break inattention ?
Are household or firms’ survey inflation expectations a policy tool?
Tool and effect• Tool: communication, speeches, !.• Effect: survey expected inflation, "#• Control: independent of other policies $• Transmission: from expected to actual inflation: "
• Question: %"%! =
%"%"# × (%"#
%! )
The three hypotheses!"#$
"% &= ((*+, #, %)
A. Is communication revealing future policy?B. Is communication revealing fundamental information?C. Is communication moving expectations independently?
The three hypotheses!"#$
"% &= ((*+, #, %)
A. Is communication revealing future policy?B. Is communication revealing fundamental information?C. Is communication moving expectations independently?• Evidence from forward guidance and policy dates, using financial markets, points to a lot of B, some A.• Spirits? Hyperinflation evidence, propaganda danger.
Two more issuesHow far away? fine tuning and anchoring
!"#$%&'
"($ )
Which inflation component? as #$ = #$+ + #$&
-"#$%&+
"($ )
Working backwards: transitory
Inflation Expectations as a Policy Tool? 11
Although consumers’ inflation expectations appear to display excess sensitivity to price changes of products in their consumption baskets, consumer prices are not equal in influencing inflation expectations. For example, Harris et al. (2009), Coibion and Gorodnichenko (2015a), Wong (2015), and others find that U.S. consumers are sensitive to gasoline prices above and beyond what is justified by the share of expenditures on gasoline.10 Panel A of Figure 2 illustrates this excess sensitivity of U.S. household inflation expectations relative to professional forecasters by plotting the two against the level of gasoline prices. There is a striking correlation between movements in the level of gasoline prices and the households’ inflation expectations. On the other hand, the relationship between gasoline prices and predictions of professional forecasters is much weaker. The same pattern holds in the euro area, as illustrated in Panel B of Figure 2.11
Relatedly, food prices also appear to have a disproportionately significant effect on inflation expectations of households (e.g., Clark and Davig 2008). Coibion and Gorodnichenko (2015b) document that Ukrainian households’ and firms’ inflation expectations react strongly to changes in the exchange rate of the hryvnia (Ukrainian currency) and the U.S. dollar. Afrouzi et al. (2018) document a similar finding in Iran.
10 Central bankers are aware of this sensitivity. Yellen (2016): “[T]he longer-run measure of inflation
expectations from the Michigan Survey has historically exhibited some sensitivity to fluctuations in current gasoline prices…” and “[A] monthly survey conducted by the Federal Reserve Bank of New York shows a noticeable decline over the past two years in household expectations for inflation three years ahead. However, these readings on shorter-term expectations may also be influenced by current gasoline prices.” Carney (2013) made a similar observation, “[W]e’ve seen a bit in the past when you have a coincident survey [of the general public’s inflation expectations] with something as obvious and important to people as energy prices move, you get these spikes.”
11 One would expect a weaker relationship between gas prices and household inflation expectations in the euro area than in the U.S. for at least two reasons. First, gasoline taxes are much higher in Europe, so a $1 increase in oil leads to a smaller percentage increase in gasoline prices in Europe than in the U.S. In addition, diesel is much more common in Europe than the U.S. (as is public transportation), making the price of basic gasoline less of a common price signal to households than in the U.S.
Figure 2 Household Inflation Expectations and Gasoline (Petrol) Prices
Panel A: United-States Panel B: Euro Area
Notes: The figure reports time series of inflation expectations of households and professional forecasters as well as gasoline (petrol) prices. All series are linearly detrended.
-2-1
01
2ga
solin
e pr
ice,
$ p
er g
allo
n (d
etre
nded
)
-4-2
02
4in
flatio
n ex
pect
atio
ns, p
erce
nt p
er y
ear (
detre
nded
)
1990 1995 2000 2005 2010 2015
Households (MSC)SPFGasoline price (right axis)
-.20
.2.4
petro
l pric
e, e
uro
per l
iter (
detre
nded
)
-20
24
6in
flatio
n ex
pect
atio
ns, p
erce
nt p
er y
ear (
detre
nded
)
2000 2002 2004 2006 2008 2010 2012 2014 2016 2018
Households (EC)SPF (ECB)Petrol price (right)
Anchoring: success of last 10 years-2
-10
12
34
5
2000m1
2001m1
2002m1
2003m1
2004m1
2005m1
2006m1
2007m1
2008m1
2009m1
2010m1
2011m1
2012m1
2013m1
2014m1
2015m1
2016m1
2017m1
USA
01
23
45
2000m1
2001m1
2002m1
2003m1
2004m1
2005m1
2006m1
2007m1
2008m1
2009m1
2010m1
2011m1
2012m1
2013m1
2014m1
2015m1
2016m1
2017m1
UK
-10
12
34
2000m1
2001m1
2002m1
2003m1
2004m1
2005m1
2006m1
2007m1
2008m1
2009m1
2010m1
2011m1
2012m1
2013m1
2014m1
2015m1
2016m1
2017m1
Eurozone
Infl. Cons.Prof.
Source: Miles, Panizza, Reis, Ubide, 2017
Fine tuning: sticky information key
0 10 20
-0.2
-0.15
-0.1
-0.05
0Inflation
0 10 20-0.2
0
0.2
FFR
0 10 20
-0.15
-0.1
-0.05
0
0.05
Infl Expect (1Q)
0 10 20
-0.15
-0.1
-0.05
0
0.05
Infl Expect (4Q)
0 5 10 15 20
-0.2
-0.1
0
Inflation Forecast Errors (1Q)
0 5 10 15 20
-0.3
-0.2
-0.1
0
Inflation Forecast Errors (4Q)
Figure 1: VAR Impulse Responses to a Monetary Shock and Conditional Forecast Errors. Upper graphs:
Impulse response functions to a monetary policy shock identified with sign restrictions. Lower graphs: The
implied one-quarter-ahead (left plot) and four-quarters-ahead (right plot) inflation forecast errors conditional
on the monetary shock. Solid lines denote posterior median responses. Shaded areas denote the 70 percent
posterior credible set. All numbers are annualized and in percent.
plot) inflation forecast errors conditional on the monetary shock. The gray areas denote the
70 percent posterior credible sets and the solid line the posterior median. Three facts have
to be emphasized. First, inflation forecast errors conditional on monetary shocks are fairly
persistent. In the aftermath of a monetary tightening, the lower graphs of Figure 1 show that
the posterior median (the solid line) of the one-quarter-ahead and four-quarters-ahead inflation
forecast errors are larger than zero for almost five years. The 70 percent posterior upper bound
for these forecast errors stays in negative territory for at least three years. Second, inflation
expectations barely move immediately after a monetary shock. Third, the responses of both
inflation and inflation expectations to monetary shocks exhibit a great deal of persistence, with
a half life20 exceeding 20 quarters. This last fact suggests that inflation expectations remain
disanchored for a few years after a monetary contraction.
While, as we shall see, the DIM can explain large and persistent conditional forecast errors
through signaling e§ects, perfect information models cannot. Indeed, the first fact is a conun-
20Half life is defined as the number of quarters after the initial shock it takes for the largest e§ect of a shockto reduce to half.
18
0 10 20
-0.2
-0.15
-0.1
-0.05
0Inflation
0 10 20-0.2
0
0.2
FFR
0 10 20
-0.15
-0.1
-0.05
0
0.05
Infl Expect (1Q)
0 10 20
-0.15
-0.1
-0.05
0
0.05
Infl Expect (4Q)
0 5 10 15 20
-0.2
-0.1
0
Inflation Forecast Errors (1Q)
0 5 10 15 20
-0.3
-0.2
-0.1
0
Inflation Forecast Errors (4Q)
Figure 1: VAR Impulse Responses to a Monetary Shock and Conditional Forecast Errors. Upper graphs:
Impulse response functions to a monetary policy shock identified with sign restrictions. Lower graphs: The
implied one-quarter-ahead (left plot) and four-quarters-ahead (right plot) inflation forecast errors conditional
on the monetary shock. Solid lines denote posterior median responses. Shaded areas denote the 70 percent
posterior credible set. All numbers are annualized and in percent.
plot) inflation forecast errors conditional on the monetary shock. The gray areas denote the
70 percent posterior credible sets and the solid line the posterior median. Three facts have
to be emphasized. First, inflation forecast errors conditional on monetary shocks are fairly
persistent. In the aftermath of a monetary tightening, the lower graphs of Figure 1 show that
the posterior median (the solid line) of the one-quarter-ahead and four-quarters-ahead inflation
forecast errors are larger than zero for almost five years. The 70 percent posterior upper bound
for these forecast errors stays in negative territory for at least three years. Second, inflation
expectations barely move immediately after a monetary shock. Third, the responses of both
inflation and inflation expectations to monetary shocks exhibit a great deal of persistence, with
a half life20 exceeding 20 quarters. This last fact suggests that inflation expectations remain
disanchored for a few years after a monetary contraction.
While, as we shall see, the DIM can explain large and persistent conditional forecast errors
through signaling e§ects, perfect information models cannot. Indeed, the first fact is a conun-
20Half life is defined as the number of quarters after the initial shock it takes for the largest e§ect of a shockto reduce to half.
18
Source: Melosi, 2018
Inattention says go beyond average46 Inflation and the Great Recession
Figure 2.14 Dispersion of Inflation expectations in the US
A: Consumers, 12 months
−2
0
2
4
6
8
10
1985
m1
1986
m1
1987
m1
1988
m1
1989
m1
1990
m1
1991
m1
1992
m1
1993
m1
1994
m1
1995
m1
1996
m1
1997
m1
1998
m1
1999
m1
2000
m1
2001
m1
2002
m1
2003
m1
2004
m1
2005
m1
2006
m1
2007
m1
2008
m1
2009
m1
2010
m1
2011
m1
2012
m1
2013
m1
2014
m1
2015
m1
2016
m1
2017
m1
B: Professional forecasters, 12 months
−2
0
2
4
6
8
10
1985
m1
1986
m1
1987
m1
1988
m1
1989
m1
1990
m1
1991
m1
1992
m1
1993
m1
1994
m1
1995
m1
1996
m1
1997
m1
1998
m1
1999
m1
2000
m1
2001
m1
2002
m1
2003
m1
2004
m1
2005
m1
2006
m1
2007
m1
2008
m1
2009
m1
2010
m1
2011
m1
2012
m1
2013
m1
2014
m1
2015
m1
2016
m1
2017
m1
Source: Miles, Panizza, Reis, Ubide, 2017Coibion Gorocnichenko, 2012
My interpretation1. Moving “animal spirits” is very hard, communicating
fundamentals and policy is unavoidable, 2. Households/firms focus on transitory, must extract
signal on permanent, so look at more moments.3. Inattention makes anchoring easier but fine tuning
harder. Compare with food labelling…
Monetary policy as a stable unit of account: success.Monetary policy as stabilization policy: harder.
Conclusion
Conclusion• Must take survey inflation expectations seriously as data.
• Shift focus, resources, and policy attention to impact of measures on surveys of inflation.
• Effects of policy: limited knowledge, but duty to inform about fundamentals and policies, better design communication.
• More evidence of success as an anchor.