Discussion of Why Has Inflation Remained So Low for So Long? By - - PowerPoint PPT Presentation

Discussion of “Why Has Inflation Remained So Low for So Long?” By Robert King (with many comments that are only partly motivated by Bob’s paper)

October 15, 2016

Jeff Fuhrer, EVP, Federal Reserve Bank of Boston

1

 Great Recession

produced an enormous

• utput/employment gap

 Yet inflation fell only a

little

 This may be the larger

puzzle

 The quick and dirty

answer: Long-run expectations remained well-anchored

 “Anchored expectations”

doesn’t mean “inflation will not budge, even during a huge recession”

2

• 8
• 6
• 4
• 2

2 4 6 2007:Q3 2007:Q4 2008:Q1 2008:Q2 2008:Q3 2008:Q4 2009:Q1 2009:Q2 2009:Q3 2009:Q4 2010:Q1 2010:Q2 2010:Q3 2010:Q4 2011:Q1 2011:Q2 2011:Q3 2011:Q4 Output gap Headline inflation Core inflation

 Compared to the 2% target

3

95 100 105 110 115 120 125 2007:Dec 2008:Apr 2008:Aug 2008:Dec 2009:Apr 2009:Aug 2009:Dec 2010:Apr 2010:Aug 2010:Dec 2011:Apr 2011:Aug 2011:Dec 2012:Apr 2012:Aug 2012:Dec 2013:Apr 2013:Aug 2013:Dec 2014:Apr 2014:Aug 2014:Dec 2015:Apr 2015:Aug 2015:Dec 2016:Apr 2016:Aug Index, Dec. 2007-100 2% price line, core PCE Core PCE Total PCE

Cumulative 4-5% gap

4

1 1.2 1.4 1.6 1.8 2 2.2 2011:Q4 2012:Q1 2012:Q2 2012:Q3 2012:Q4 2013:Q1 2013:Q2 2013:Q3 2013:Q4 2014:Q1 2014:Q2 2014:Q3 2014:Q4 2015:Q1 2015:Q2 2015:Q3 2015:Q4 2016:Q1 2016:Q2 Core Pce SPF

 RMSE for inflation forecasts, recent history  Current core PCE: 1.7% (12-month)  0.3 ppt below target is well within forecast accuracy  We’ll come back to this.

5

RMS Forecast errors, SPF Core PCE, 2007-present Horizon RMSE Current quarter 0.50 t+1 0.58 t+2 0.60 t+3 0.60 t+4 0.58

 The output/employment gap is not yet zero  Real marginal cost/labor share remains low, even

accounting for trend

 So inflation has been a little low  That’s a pretty good Squawk Box explanation, but you

might want something more substantial…

6

 Depends on what modeling framework one uses

 Accelerationist Phillips curve?  New-Keynesian Phillips curve?  So-called “anchored expectations” “model”?

7

 Accelerationist Phillips curve (toy version)

 Defines the change in inflation; CB influences changes so as to

attain its target

 Once inflation is below the central bank’s target, must have a

period of below-NAIRU unemployment to get it back up

 Could be that’s why inflation is stubbornly low today  But look at this model’s forecast for inflation during recession:

8 of 15

* 1

( )

t t t t t

U U π π β ε

−

= − − +

• 8
• 6
• 4
• 2

2 4

2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 Flatter PC Estimated PC Toy PC

 Inflation importantly linked to “fundamental inflation”—

the discounted expectations of output or real marginal cost

 Papers differ in their claims for the success of the NKPC

in explaining inflation during the GR

 Del Negro et al find a model that fits the data fairly well  Next slides present some estimates of fundamental inflation

 NB: Expectations are perfectly anchored in this model

9

1

(1 )

F t t t t F j t t t j j

E s π ωπ ω π µ π λ β

− ∞ + =

= + − + = ∑

2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Year

• 12
• 10
• 8
• 6
• 4
• 2

2 4

Percent

Core PCE Fundamental inflation

Estimates of fundamental inflation

Various estimation samples, discount factors

Samples include or exclude the 1980s, the Great Recession Discount factors from 0.5 to 0.98

10

 Method: Like King-

Watson (2012). 8- variable VAR (includes long-run

• infl. expectations,
• utput, wages, ulc, C,

I, ff)

 Add an identity

defining fundamental inflation, using VAR forecasts of labor share

 Trend in labor share is

a serious problem

 King and Watson point

this out

Labor share not detrended

1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018

Year

• 3
• 2
• 1

1 2 3 4 5 6

Percent

Core PCE Fundamental inflation

Estimates of fundamental inflation

Various estimation samples, discount factors

Samples include or exclude the 1980s, the Great Recession Discount factors from 0.5 to 0.98

11

Labor share detrended



Use detrended labor share



Where does trend come from?



Most of these estimates of “fundamental” inflation have a hard time explaining recent inflation



They generally suggest that inflation should have been even lower than it has been



Similar results using output gap instead of L share

 The simple version  As output gap disappears, inflation goes to long-run

expectation

 Somewhat like NKPC, but only LR expectations are explicit

 Important that the long-run expectation remain

“anchored”—i.e. fixed at CB target

 If not, inflation could settle somewhere other than CB

target, even as output goes to equilibrium

 How does a CB make sure that is the case?

12

* *

( ) ; ?

LR LR t t t t t CB

a y y π π ε π π = + − + =

1985 1990 1995 2000 2005 2010 2015 2020

Year

0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Core PCE Fitted value Forecast

 They kinda do

 Long-run

expectations capture the trend

 The gap captures

some of the up and down prior to and during the recession

 Puzzle for this model:

Output gap is probably nearly closed, possibly turning positive in the coming years

 But inflation is still

below 2%

 But not a huge puzzle

13

* 1 1 2 2 1 2

(1 ) ( )

LR t t t t t t t

l l l l a y y π π π π ε

− −

= + + − − + − +

Long-run expectations = FRB/US PTR; output gap uses CBO estimate of potential GDP; estimation from 1998:1-2016:2

 The model says that once we get output to equilibrium,

inflation equals long-run expected inflation

 But not explicitly because the central bank is acting to move

inflation toward the target

 In fact, the CB could just target output, and inflation would

always return to the long-run expectation

 The model above behaves just fine with NO policy

response to inflation—that’s a little different (certainly not true for standard models, e.g. with NKPC or OKPC)

 Not just an academic point—this is about how CB’s ensure

that inflation returns target, a central issue today.

14

* 1 1 1 1 *

( ); * ( 2) 2 * ( )

LR LR t t t t t t t t t t t y t t

l y y y Ey f E f a y y π π π β π π σ π π

− + +

= + + − = = − − − = + + −

 Bob identifies a few surprises (possibly puzzles)

 2008 decline; return to 2% in 2011; below-2% inflation recently

 How big are these surprises/puzzles? Compare to history.

15

08:Q1 08:Q3 09:Q1 09:Q3 10:Q1 10:Q3 11:Q1

Realization date

• 1.5
• 1
• 0.5

0.5 1 1.5

1-qtr. 4-qtr.

Tealbook forecast errors, core CPI, 2008-2010

Forecast-actual 1985 1990 1995 2000 2005 2010 2015

Realization date

• 2.5
• 2
• 1.5
• 1
• 0.5

0.5 1 1.5 2

1-qtr. 4-qtr.

Tealbook forecast errors, core CPI, 1990-2010

Forecast-actual

Not so large, nor too one-sided.



Survey-based PC does OK during recent period



Suggests that a better understanding

• f

expectations may improve

• ur imperfect

understanding

• f inflation

16

,10 ,1 ,1 *

(1 ) ( )

SPF yr SPF q SPF q t t t t t

a a b U U π π π = + − − −

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Year

0.5 1 1.5 2 2.5 3 Core CPI inflation In-sample fit Out-of-sample forecast

Phillips curve model with SPF expectations (long- and short-run)



Intrinsic persistence in expectations formation (macro evidence-Fuhrer 2015)



That is, (short-run) expectations add persistence beyond what is in the series for which they are forming expectations



Micro-data on expectations display a specific kind of “intrinsic persistence”- Fuhrer 2016



The source: Forecasters and households keep their forecasts close to the lagged central tendency of forecasts



What’s the evidence? (SPF, Michigan, Euro SPF)

17

Slope = -0.6 p-value = 0.000

Dependent variable: Revision from t-1 to t for forecast period Lagged discrepancy

t+1 t+2 t+3 t+1 t+2 t+3

• 0.58

(0.000)

• 0.56

(0.000)

• 0.54

(0.000)

• 0.53

(0.000)

• 0.61

(0.000)

• 0.61

(0.000) 0.01 (0.648) 0.03 (0.019) 0.05 (0.000) Other forecast controls N N N Y Y Y Observations 3274 3257 3180 3029 3017 2960

1, 1 1| 1 i Median t t t t

π π

+ − + −

−

2, 1 2| 1 i Median t t t t

π π

+ − + −

−

3, 1 3| 1 i Median t t t t

π π

+ − + −

−

1 i t

π −

ith forecaster’s revision

Discrepancy between ith lagged forecast and lagged median Scatter plot relating forecast revisions to lagged discrepancy between individual and median forecast

 Intuition is simple:  During Great Recession,

expectations inertia tempered the decline in inflation

 See this in the survey-

based PC

 Similarly, during recovery,

expectations inertia slows the increase toward the long-run target

 Simple macro model with

this kind of explicit expectations inertia illustrates these effects:

18

8 16 24 32 40 48

Quarters

0.5 1 1.5 2

Simulated, with expectations inertia Acutal core PCE (12-qtr.)

Inflation during recession and recovery with inertial expectations

 Inflation dynamics are still not well understood, especially in

low-inflation environments

 Although forecast misses/model failures this time are not particularly

large

 Some serial correlation of errors, but not striking  Thought errors aren’t huge, we’d like to hit our target!  Simple old-style Phillips curves can’t explain both episodes

(recession and recovery)

 Neither do NKPC’s (by my estimates)  Anchored expectations—sort of do (not perfectly)  But the theoretical underpinnings, and the links to real-world price-

setting, are somewhat suspect (to me)

 Alternative models of expectations (perhaps captured well by

surveys), both short- and long-term, may help

19