Discussion of ``IS EURO AREA LOWFLATION HERE TO STAY?'' by ... · OVERVIEW Key question Should recent persistence of low in ation translate intopermanentlylower in ation expectations?

Discussion of“IS EURO AREA LOWFLATION

HERE TO STAY?”by Stevens and Wauters

Elmar Mertens

Deutsche Bundesbank

The discussion and analysis presented here does not necessarily reflectthe views of the Deutsche Bundesbank or the Eurosystem

NBB conference, Brussels, October 2018

OVERVIEW

Key question

Should recent persistence of low inflation translateinto permanently lower inflation expectations?

Approach

• Trend-cycle decomposition for inflation(w/time-varying parameters)

• Surveys added to model’s measurement equation

Findings

1 With survey data: inflation trend below but close to 2%

2 Without survey data: trend estimates falling to 1.5%

3 High degrees of information rigidity embedded in surveys

OVERVIEW

Key question


Approach



Findings




OVERVIEW

Key question


Approach



Findings




OVERVIEW

Key question


Approach



Findings




PAUL VOLCKER’S PERSPECTIVEWashington Post, October 24 2018

On the FOMC’s inflation objective

They made up the 2 percent number . . .

I get upset when I hear them fighting over whether1.75 percent is enough inflation.

On the importance of remaining vigilant:

Two percent inflation isn’t going to kill us . . .

But be careful of 2.3 percent being ok and then theysay let’s let it go to 3 percent.

AGENDA

1 Survey-based inflation trend estimates

2 Sticky information state space

3 State dependent stickiness

TREND INFLATION

Beveridge-Nelson trend in inflation

τt ≡ limk→∞

Etπt+k

• univariate: Stock & Watson “UCSV” (2007, JMCB)

• multivariate, common trend: Mertens (2016, REStat)

Survey

• Ftπt+h = Etπt+h + zt+hwhere zt measures deviations from RE

• Weak rationality: Etzt+∞ = 0 (Grant & Thomas, 1999)

• τt = limk→∞Et (Ft+kπt+k+h) = limk→∞Etπt+k+h• τt is common trend of inflation and surveys

Holds in Stevens-Wauters model

SURVEYS AND TREND INFLATION


τt ≡ limk→∞

Etπt+k



Survey data







τt ≡ limk→∞

Etπt+k



Survey trend







τt ≡ limk→∞

Etπt+k



Survey trend





EURO-AREA INFLATION TRENDUnivariate UCSV Trend (red), 12m inflation data (blue)

1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016

-1

0

1

2

3

4

5

As in Mertens (2016): Deviations from trend as VAR

EURO-AREA INFLATION TRENDUnivariate UCSV Trend (red)

1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016

-1

0

1

2

3

4

5


EURO-AREA INFLATION TRENDUnivariate UCSV Trend (red), Common Trend (black) w/surveys

1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016

-1

0

1

2

3

4

5


COMMON TREND PERSPECTIVE: TAKE AWAYS

• Cointegration between surveys and realizedinflation useful to exploit

How much to gain from the specific Phillips-curvemodel for gap inflation used here?

• U.S.: Survey-based trend estimates laggedinflation-based estimates in 1980s/90s

• Forecasts centered around inflation-based trendestimates during 1980s/90s in the U.S. would haveworked better

INFLATION TREND LEVELS MERTENS (2016, RESTAT)

INFTRM (red): inflation-data-based, SRV (black): survey-based

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 20150

2

4

6

8

10

12SRVINFTRM

COMMON TREND PERSPECTIVE: TAKE AWAYS

• Cointegration between surveys and realizedinflation useful to exploit

How much to gain from the specific Phillips-curve modelfor gap inflation used here?

• U.S.: Survey-based trend estimates laggedinflation-based estimates in 1980s/90s

• Forecasts centered around inflation-based trendestimates during 1980s/90s in the U.S. would haveworked better

AGENDA




INFORMATION RIGIDITIES AND FORECASTINGa.k.a. Stevens-Wauters “forecast smoothing”

Coibion & Gorodnichenko (2012 JPE, 2015 AER)

Ftπt+h = (1 − ξ)Etπt+h + ξFt−1πt+h

Encompassing Mankiw-Reis stickiness, noisy information,Sims-Mackowiak-Wiederholt rational inattention

Stevens & Wauters (2018)

Ftπt+h = (1 − ξt)Etπt+h + ξtFt−1πt+h−1

citing use of rolling-event forecasts

Mertens & Nason (2018)

Ftπt+h = (1 − ξt)Etπt+h + ξtFt−1πt+h

State space generates RE and SI forecastsfor any horizons and events

(see appendix)











(see appendix)











(see appendix)

STATE SPACE MODEL FOR INFLATIONMertens & Nason (2018); applicable also to Stevens & Wauters

Inflation dynamics

πt = CXt Xt =[τπt , πt − τπt . . .

]Xt = AXt−1 +Bwt Etπt+h = C AhXt


Inflation dynamics



Stevens-Wauters vs Mertens-Nason:• Similar: Reduced form, independent from survey dynamics

• New: Phillips curve with unemployment-rate gap

Mertens-Nason not a univariate inflation process

• Inflation driven by multiple state variables whoseestimates are informed by SPF

• Given information from SPF, how much to be gainedfrom reduced-form PC?


Inflation dynamics



Stevens-Wauters vs Mertens-Nason:• Similar: Reduced form, independent from survey dynamics

• New: Phillips curve with unemployment-rate gap

Mertens-Nason not a univariate inflation process

• Inflation driven by multiple state variables whoseestimates are informed by SPF

• Given information from SPF, how much to be gainedfrom reduced-form PC?

COMMENTS ON STEVENS-WAUTER INFLATION MODEL

Not a model of forward-looking inflation

• No effects from surveys onto inflation

• Cannot speak to effects of information rigiditieson inflation(e.g. Coibion & Gorodnichenko 2015, AEJM)

• Surveys affect trend (and gap) estimates

• But: allows to derive recursive state space. Use it!

Other comments

• Embed ECB inflation target (Chan et al bounds)?

• Import prices not relevant for trend identification,only for variance decomposition

• Specification choices: no trend SV, ρπt > 0, etc . . .







Other comments










Other comments










Other comments










Other comments










Other comments




EURO-AREA INFLATION TRENDSmoothed trend estimates: Univariate UCSV (red), Mertens-Nason (black)

1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016

-1

0

1

2

3

4

5

AGENDA




SI WEIGHT ξt EURO AREA

Mertens-Nason w/Stevens-Wauters data, smoothed (black), filtered (red)

1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

No significant movements since 2001 (see appendix)Survey data available since 1998:Q4

INFLATION PERSISTENCE AND SI WEIGHTFiltered estimates from Mertens & Nason (2018), U.S. data.

STICKINESS

1970 1975 1980 1985 1990 1995 2000 2005 2010 2015

0

0.2

0.4

0.6

0.8

SENSITIVITY OF RE FORECASTS TO SHOCKS

1970 1975 1980 1985 1990 1995 2000 2005 2010 2015

0

0.5

1

1.5

2

2.5

3

h = 1

h = 5

Link between inflation persistence and attention?

CONCLUSION

The question was . . .


The answer is . . .

No, at least not whentrends estimates are generated from surveys

Thoughts

• Could better motivate some model choices

• Causes of survey stickiness?

• Relative constancy of euro-area stickiness indicative ofsuccesful anchoring?

Very nice paper!

CONCLUSION



The answer is . . .


Thoughts




Very nice paper!

CONCLUSION



The answer is . . .


Thoughts




Very nice paper!

APPENDIX

INFLATION GAP AR(1) PARAMETER BACKUP


1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

TREND SHOCK VOL BACKUP


1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016

0

0.2

0.4

0.6

0.8

1

1.2

1.4

GAP SHOCK VOL BACKUP


1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016

0

0.5

1

1.5

2

2.5

3

DIFFERENCES IN SI WEIGHT OVER TIME ξt EURO AREA

ξt|T − ξ2001|T

2000 2002 2004 2006 2008 2010 2012 2014 2016

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

Inference based on joint uncertainty between ξt and ξ2001

RECURSIVE STATE SPACE WITH SURVEYSMertens & Nason (2018); applicable also to Stevens & Wauters

Inflation dynamics



Sticky-information survey states

Ftπt+h = C FtXt+h

FtXt+h = Ah FtXt

FtXt = (1 − ξt−1)Xt + ξt−1Ft−1Xt

Can construct survey forecastsfor every event and horizon

(w/AUM forecasts of time-varying parameters)


Inflation dynamics




Ftπt+h = C FtXt+h

FtXt+h = Ah FtXt

FtXt = (1 − ξt−1)Xt + ξt−1A Ft−1Xt−1




Inflation dynamics


]Xt = AtXt−1 +Bwt Etπt+h = C AhXt


Ftπt+h = C FtXt+h

FtXt+h = Aht FtXt

FtXt = (1 − ξt−1)Xt + ξt−1A Ft−1Xt−1




Inflation dynamics


]Xt = AtXt−1 +Bwt Etπt+h = C AhXt


Ftπt+h = C FtXt+h

FtXt+h = Aht FtXt

FtXt = (1 − ξt−1)Xt + ξt−1At−1 Ft−1Xt−1



CONCLUSION



The answer is . . .


Thoughts


• Causes of forecast smoothing?

• Is relative constancy of euro-area stickiness indicative ofsuccesful anchoring?

Very nice paper!

Discussion of ``IS EURO AREA LOWFLATION HERE TO STAY?'' by ... · OVERVIEW Key question Should recent persistence of low in ation translate intopermanentlylower in ation expectations?

Documents