Empirical Assessment of the Phillips Curve · 2018. 2. 5. · Slope of Phillips curve: Causal...

EMPIRICAL ASSESSMENT OF THE PHILLIPS CURVE

Emi Nakamura Jón Steinsson

Columbia University

January 2018

Nakamura-Steinsson (Columbia) Phillips Curve January 2018 1 / 55

BRIEF HISTORY OF THE PHILLIPS CURVE

Phillips 58 points out empirical relationship between wage inflation

and unemployment in UK 1861-1957

Samuelson-Solow 60 popularize idea in US


INFLATION AND UNEMPLOYMENT IN THE UK

1958] UNEMPLOYMENT AND MONEY WAGE RATES 285

year, an increase of 7 * 6 per cent. in 1900 and in 1910, and an increase of 7 0 per cent. in 1872. In no other year between 1861 and 1913 was there an increase in import prices of as much as 5 per cent. If the hypothesis stated above is correct the rise in import prices in 1862 may just have been sufficient to start up a mild wage-price spiral, but in the remainder of the period changes in import prices will have had little or no effect on the rate of change of wage rates.

U 0

cf.

E? 2 - 0 0~~~~

cl)

? -2 0

. ~~~~~~~~~~*

.-40 1 2 3 4 6 $ 7 8 9 10 11 Unemptoyment, %.

Fig,t. 1861 - 1913

A scatter diagram of the rate of change of wage rates and the percentage unemployment for the years 1861-1913 is shown in Figure 1. During this time there were 61 fairly regular trade cycles with an average period of about 8 years. Scatter diagrams for the years of each trade cycle are shown in Figures 2 to 8. Each dot in the diagrams represents a year, the average rate of change of money wage rates during the year being given by the scale on the vertical axis and the average unemployment during the year by the scale on the horizontal axis. The rate of change of money wage rates was calculated from the index of hourly wage rates constructed by Phelps Brown and Sheila Hopkins,' by expressing the first central difference of the index for each year as a percentage of the index for the same year. Thus the rate of change for 1861 is taken to be half the difference between the index for 1862 and the index for 1860 expressed as a percentage of the index

IE. HI. Phelps Brown and Sheila Hopkins, "'The Course of Wage Rates in Five Countries, 1860-1939," Oxford Economic Papers, June, 1950.

This content downloaded from 128.59.165.222 on Fri, 26 Jan 2018 16:23:32 UTCAll use subject to http://about.jstor.org/terms

Source: Phillips (1958)


INFLATION AND UNEMPLOYMENT IN THE UK 286 ECONOMICA [NOVEMBER

10

@i 10 X.

*-,Curve fitted to 1861-1913 data

6-

IV

4 64 63

c 2 E: 2- t-t 2 > ~~~~~62

-2 z ~~~~~~~~~67

I I I * a I I I I I B 0 0 1 2 3 4 5 6 7 8. 9 10 11

Unemployment, %.

Fig.2. 1861 - 1868

2 Curve fitted to 1861-1913 data 72

6-6 73

c) 0

3: 4l Z'% 70

C

0 Unmpoyen,74.

Fig.3. 1868 -1879 ~ ~ ~ 7





286 ECONOMICA [NOVEMBER

10

@i 10 X.

*-,Curve fitted to 1861-1913 data

6-

IV

4 64 63

c 2 E: 2- t-t 2 > ~~~~~62

-2 z ~~~~~~~~~67

I I I * a I I I I I B 0 0 1 2 3 4 5 6 7 8. 9 10 11

Unemployment, %.

Fig.2. 1861 - 1868

2 Curve fitted to 1861-1913 data 72

6-6 73

c) 0

3: 4l Z'% 70

C

0 Unmpoyen,74.

Fig.3. 1868 -1879 ~ ~ ~ 7




INFLATION AND UNEMPLOYMENT IN THE UK 288 ECONOMICA [NOVEMBER

C

t1:

10-

-f1C Curve fitted to 1861 -1913 data

+' 6 \

06 0) 3 4 \9

o 90 E 87 86 -0 0 92 93 0) C -2

0 1 2 3 4 5 6 7 8 9 10 I

cr. Unemptoyment, %.

Fig.5. 1886- 1893

># 10

* \ Curve fitted to 1861-1913 data

6-

(U4

0 0 no4- \

E 0 @ ~~~~~01 02 os 04 94 93

X-2-

"o -4J 0 1 2 3 4 5 6 7 8 9 10 11

Unemployment, 1.

Fig.6. 1893 -1904





288 ECONOMICA [NOVEMBER

C

t1:

10-

-f1C Curve fitted to 1861 -1913 data

+' 6 \

06 0) 3 4 \9

o 90 E 87 86 -0 0 92 93 0) C -2

0 1 2 3 4 5 6 7 8 9 10 I

cr. Unemptoyment, %.

Fig.5. 1886- 1893

># 10

* \ Curve fitted to 1861-1913 data

6-

(U4

0 0 no4- \

E 0 @ ~~~~~01 02 os 04 94 93

X-2-

"o -4J 0 1 2 3 4 5 6 7 8 9 10 11

Unemployment, 1.

Fig.6. 1893 -1904




US MACRO POLICY IN THE 1960S

Phillips curve viewed as a menu of options

Policy makers can lower unemployment if they are

willing to tolerate more inflation


INFLATION AND UNEMPLOYMENT IN THE USU.S. in the 1960’s

46

0.00

2.00

4.00

6.00

8.00

10.00

12.00

14.00

16.00

2.00 4.00 6.00 8.00 10.00 12.00

Infla

tion Ra

te

Unemployment Rate

19601965

1969


FRIEDMAN AND PHELPS MAKE A PREDICTION

Friedman 68 and Phelps 67:

Policymakers cannot exploit a stable Phillips curve forever

Workers will demand wage increases in excess of expected inflation

As inflation rises, expectations of inflation will rise

Changes in expected inflation will shift the Phillips curve


INFLATION AND UNEMPLOYMENT IN THE USU.S. in the 1960’s

46

0.00

2.00

4.00

6.00

8.00

10.00

12.00

14.00

16.00

2.00 4.00 6.00 8.00 10.00 12.00

Infla

tion Ra

te

Unemployment Rate

19601965

1969


FRIEDMAN AND PHELPS WERE RIGHT!Friedman and Phelps Were Right!!

9

0.00

2.00

4.00

6.00

8.00

10.00

12.00

14.00

16.00

2.00 4.00 6.00 8.00 10.00 12.00

Infla

tion Ra

te

Unemployment Rate

19601965

1969

1980


MODERN PHILLIPS CURVE

πt = βEtπt+1 + κ(yt − ynt ) + ηt

Three drivers of inflation:

Expected inflation: Etπt

Output relative to potential: yt − ynt

Cost-push shocks: ηt

Specific form above based on Calvo 83 sticky-price assumptions

But details may vary (e.g., sticky information yields Et−1πt )


ESTIMATING SLOPE OF THE PHILLIPS CURVE


Object of interest: Slope coefficient κ

How much does an increase in “demand” / “tightness” / “output gap”

affect inflation

Tricky identification issues:

Expected inflation unobserved

“Natural rate of output” (i.e., supply shocks) unobserved

Cost push shocks (e.g., variation in desired markups) unobserved

All three may cause omitted variables bias


EXPECTED INFLATION

Pre Friedman/Phelps Phillips curve: Change in output gap needed

to change inflation

Same is true for accelerationist Phillips curve

(i.e., Phillips curve with adaptive expectations)

πt = πt−1 + κ(yt − ynt )

Sargent 82: Hyperinflations end abruptly with little or no output cost

Clear violation of aforementioned Phillips curves


GERMAN HYPERINFLATION

9

Sargent (1982): The Ends of Four Big Inflations

Source: Sargent (1982)


EXPECTED INFLATION

In Calvo model, perfectly credible, unexpected disinflation can occurwithout any affect on output gap

Expected inflation does all the work

Theoretical victory: Potential explanation for Sargent facts

Empirical headache:

Movements in inflation potentially completely unrelated to output gap

Even if output gap moves during disinflation, not clear what fraction

of disinflation was due to shift in expected inflation

Measurement of expected inflation crucial but hard


SUPPLY SHOCKS

Slope of Phillips curve: Causal effect of output gap on inflation

But output gap is not directly observable

Ideal experiment:

Shifts in output holding potential output and expected inflation constant

Shifts in potential output (or natural rate of unemployment):

Shift the Phillips curve

Potentially lead to negative correlation between output and inflation

Bias estimated slope of Phillips curve downward

Cost push shocks cause similar problems


IS THE PHILLIPS CURVE DEAD?

Phillips curve often pronounced dead

Many economists think Phillips curve is an empirical disaster

Prominent episodes:

Missing inflation in late 1990s

Missing disinflation in the Great Recession

Let’s take a closer look


STAGFLATION OF THE 1970S

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

2.0 4.0 6.0 8.0 10.0

Inflation

Unemployment

1969

1980

1961


STAGFLATION OF THE 1970S

Friedman and Phelps were right:

Phillips curve shifted up in 1970s

But why did inflation rise at higher and higher unemployment rates

as 1970s progressed?

Possible explanation: Rise in natural rate of unemployment

Baby boon generation was entering labor force in 1970s


VOLCKER DISINFLATION

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

2.0 4.0 6.0 8.0 10.0

Inflation

Unemployment

1969

1980

1983

1989

1961


MISSING INFLATION IN LATE 1990S

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0

Inflation

Unemployment

1986

1990

19922000


MISSING INFLATION IN LATE 1990S

1986-1996: “Looks good”

1997-2000: Why doesn’t inflation rise?

Maybe it was just starting to in 2000 but Fed nipped it in the bud

Maybe the Phillips curve is highly non-linear


MISSING DISINFLATION IN THE GREAT RECESSION

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0

Inflation

Unemployment

1986

1990

19922000

2003

2007

2010


MISSING DISINFLATION IN THE GREAT RECESSION

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0

Inflation

Unemployment

1990

19922000 2007

2010

20142017


IS THE PHILLIPS CURVE DEAD?

Why did inflation not fall more in Great Recession?

Actually rose in 2010 and 2011

Why is inflation not rising now?


COIBION-GORODNICHENKO 15

“Is the Phillips Curve Alive and Well after All?”

Focus on “missing disinflation” during Great Recession

Population explanations insufficient

Anchored inflationary expectations

Movements in natural rate

Flattening of the Phillips curve

New explanation:

Household inflation expectations rose in 2009-2013

If firm’s expectation the same, this can explain missing disinflation




Baseline assumptions:

Output gap measure: Unemployment rate

yt − ynt = ut

Expectations of inflation: backward looking

Etπt+1 =14

(πt−1 + πt−2 + πt−3 + πt−4)

Ignore discounting: β = 1



πt − EπBackt+1 = κut + ηt

Estimate by OLS for sample 1960Q1-2007Q4

Implicitly assuming that ηt ⊥ ut

Similar to VAR timing restriction. (Why?)

Consider whether Great Recession conforms with earlier relationship


MISSING DISINFLATION202 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS JANUARY 2015

−10

−8

−6

−4

−2

0

2

4

6

8

Infla

tion

2010 2011 2013

07Q4

08Q1

08Q2

08Q3

09Q1

09Q2

09Q309Q4

10Q1

10Q2

10Q3

10Q4

11Q1

11Q2

11Q3

11Q4

12Q1

12Q2

12Q312Q4

13Q1

−8

−6

−4

−2

0

2

4

3 4 5 6 7 8 9 10 11

Unemployment rate

Panel A. CPI in�ation and US unemployment

2007 2008 2009 2012

Panel B. CPI inflation and predicted inflation from the Phillips Curve

Actual

Backward PC

π t−

Eπ t

Bac

k

Figure 1. The Missing Disinflation

Notes: Panel A shows the scatter plot of inflation surprises ( π t − E t π t BACK ) versus unemployment rate. E t π t BACK is calculated as in equation (2). Empty circles show observations for 1960Q1–2007Q3. Filled circles show observa-tions for 2007Q4–2013Q1. The solid line shows predicted inflation surprises as a function of the unemployment rate in the linear regression. The inflation surprise for 2008Q4 is outside the range of the figure and is not reported. Panel B plots time series of the actual CPI inflation rate (annualized; solid thick line) and the CPI inflation rate pre-dicted by the Phillips curve (equation (1); dashed line) which is estimated on the 1960Q1–2006Q3 sample.

Source: Coibion and Gorodnichenko (2015)


MISSING DISINFLATION

202 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS JANUARY 2015

−10

−8

−6

−4

−2

0

2

4

6

8

Infla

tion

2010 2011 2013

07Q4

08Q1

08Q2

08Q3

09Q1

09Q2

09Q309Q4

10Q1

10Q2

10Q3

10Q4

11Q1

11Q2

11Q3

11Q4

12Q1

12Q2

12Q312Q4

13Q1

−8

−6

−4

−2

0

2

4

3 4 5 6 7 8 9 10 11

Unemployment rate

Panel A. CPI in�ation and US unemployment

2007 2008 2009 2012

Panel B. CPI inflation and predicted inflation from the Phillips Curve

Actual

Backward PC

π t−

Eπ t

Bac

k

Figure 1. The Missing Disinflation

Notes: Panel A shows the scatter plot of inflation surprises ( π t − E t π t BACK ) versus unemployment rate. E t π t BACK is calculated as in equation (2). Empty circles show observations for 1960Q1–2007Q3. Filled circles show observa-tions for 2007Q4–2013Q1. The solid line shows predicted inflation surprises as a function of the unemployment rate in the linear regression. The inflation surprise for 2008Q4 is outside the range of the figure and is not reported. Panel B plots time series of the actual CPI inflation rate (annualized; solid thick line) and the CPI inflation rate pre-dicted by the Phillips curve (equation (1); dashed line) which is estimated on the 1960Q1–2006Q3 sample.





movements during this period pushed inflation up despite the weak economy. The price of West Texas Intermediate (WTI) crude, for example, went from under 40 dollars per barrel in early 2009 to over 100 dollars per barrel in early 2011, precisely the period during which inflation was significantly higher than expected from historical Phillips curve correlations. To assess whether changing oil prices can account for the unusual inflation dynamics during this period via shifts in

07Q408Q1

08Q2

08Q3

08Q4

09Q109Q2

09Q3

09Q4

10Q1

10Q210Q310Q4

11Q111Q211Q3

11Q412Q112Q2

12Q312Q4

13Q1

07Q4

08Q1

08Q2

08Q3

09Q1

09Q2

09Q309Q4

10Q1

10Q2

10Q3

10Q4

11Q111Q2

11Q3

11Q4

12Q1

12Q2

12Q312Q4

13Q1

07Q4

08Q1

08Q2

08Q3

09Q1

09Q2

09Q3

09Q4

10Q1

10Q2

10Q3

10Q4

11Q1

11Q2

11Q3

11Q4

12Q112Q2

12Q3

12Q4

13Q1

−2 0 2 4 6

Unemployment rate, partial out oil price changes

07Q408Q1

08Q208Q3

09Q1

09Q2

09Q3

09Q410Q110Q210Q3

10Q411Q111Q211Q3

11Q4

12Q112Q2

12Q3

12Q413Q1

07Q408Q1

08Q2

08Q3

09Q1

09Q209Q3

09Q4

10Q110Q210Q310Q4

11Q1

11Q211Q3

11Q4

12Q1

12Q212Q312Q413Q1

07Q4

08Q1

08Q2 08Q3

09Q1

09Q2

09Q3

09Q4

10Q1

10Q210Q3

10Q4

11Q111Q2

11Q3

11Q4

12Q1

12Q2

12Q312Q4

13Q1

−8

−6

−4

−2

0

2

4

−8

−6

−4

−2

0

2

4

−6

−4

−2

0

2

4

−6

−4

−2

0

2

4

−8

−6

−4

−2

0

2

4

−8

−6

−4

−2

0

2

4

3 4 5 6 7 8 9 10 11

Unemployment rate3 4 5 6 7 8 9 10 11

Unemployment rate

3 4 5 6 7 8 9 10 11Unemployment rate


3 4 5 6 7 8 9 10 11

Unemployment rate

Panel A. PCE in�ation

Panel C. Core CPI in�ation Panel D. Core PCE in�ation

Panel E. SPF in�ation (CPI) forecasts Panel F. Controlling for oil prices

Panel B. GDP de�ator in�ation

π t − E

π tBA

CK

PC

E in

flatio

n

π t − E

π tBA

CK

GD

P d

efla

tor

infla

tion

π t − E

π tBA

CK

CP

I CO

RE

infla

tion

π t − E

π tSP

F

π t − E

π tBA

CK

part

ial o

ut o

il pr

ice

chan

ges

π t − E

π tBA

CK

P

CE

CO

RE

infla

tion

Figure 2. Robustness of the Missing Disinflation

Note: See notes for Figure 1 and the text for more details.






07Q408Q1

08Q2

08Q3

08Q4

09Q109Q2

09Q3

09Q4

10Q1

10Q210Q310Q4

11Q111Q211Q3

11Q412Q112Q2

12Q312Q4

13Q1

07Q4

08Q1

08Q2

08Q3

09Q1

09Q2

09Q309Q4

10Q1

10Q2

10Q3

10Q4

11Q111Q2

11Q3

11Q4

12Q1

12Q2

12Q312Q4

13Q1

07Q4

08Q1

08Q2

08Q3

09Q1

09Q2

09Q3

09Q4

10Q1

10Q2

10Q3

10Q4

11Q1

11Q2

11Q3

11Q4

12Q112Q2

12Q3

12Q4

13Q1

−2 0 2 4 6


07Q408Q1

08Q208Q3

09Q1

09Q2

09Q3

09Q410Q110Q210Q3

10Q411Q111Q211Q3

11Q4

12Q112Q2

12Q3

12Q413Q1

07Q408Q1

08Q2

08Q3

09Q1

09Q209Q3

09Q4

10Q110Q210Q310Q4

11Q1

11Q211Q3

11Q4

12Q1

12Q212Q312Q413Q1

07Q4

08Q1

08Q2 08Q3

09Q1

09Q2

09Q3

09Q4

10Q1

10Q210Q3

10Q4

11Q111Q2

11Q3

11Q4

12Q1

12Q2

12Q312Q4

13Q1

−8

−6

−4

−2

0

2

4

−8

−6

−4

−2

0

2

4

−6

−4

−2

0

2

4

−6

−4

−2

0

2

4

−8

−6

−4

−2

0

2

4

−8

−6

−4

−2

0

2

4

3 4 5 6 7 8 9 10 11


Unemployment rate



3 4 5 6 7 8 9 10 11

Unemployment rate





π t − E

π tBA

CK

PC

E in

flatio

n

π t − E

π tBA

CK

GD

P d

efla

tor

infla

tion

π t − E

π tBA

CK

CP

I CO

RE

infla

tion

π t − E

π tSP

F

π t − E

π tBA

CK

part

ial o

ut o

il pr

ice

chan

ges

π t − E

π tBA

CK

P

CE

CO

RE

infla

tion


Note: See notes for Figure 1 and the text for more details.






07Q408Q1

08Q2

08Q3

08Q4

09Q109Q2

09Q3

09Q4

10Q1

10Q210Q310Q4

11Q111Q211Q3

11Q412Q112Q2

12Q312Q4

13Q1

07Q4

08Q1

08Q2

08Q3

09Q1

09Q2

09Q309Q4

10Q1

10Q2

10Q3

10Q4

11Q111Q2

11Q3

11Q4

12Q1

12Q2

12Q312Q4

13Q1

07Q4

08Q1

08Q2

08Q3

09Q1

09Q2

09Q3

09Q4

10Q1

10Q2

10Q3

10Q4

11Q1

11Q2

11Q3

11Q4

12Q112Q2

12Q3

12Q4

13Q1

−2 0 2 4 6


07Q408Q1

08Q208Q3

09Q1

09Q2

09Q3

09Q410Q110Q210Q3

10Q411Q111Q211Q3

11Q4

12Q112Q2

12Q3

12Q413Q1

07Q408Q1

08Q2

08Q3

09Q1

09Q209Q3

09Q4

10Q110Q210Q310Q4

11Q1

11Q211Q3

11Q4

12Q1

12Q212Q312Q413Q1

07Q4

08Q1

08Q2 08Q3

09Q1

09Q2

09Q3

09Q4

10Q1

10Q210Q3

10Q4

11Q111Q2

11Q3

11Q4

12Q1

12Q2

12Q312Q4

13Q1

−8

−6

−4

−2

0

2

4

−8

−6

−4

−2

0

2

4

−6

−4

−2

0

2

4

−6

−4

−2

0

2

4

−8

−6

−4

−2

0

2

4

−8

−6

−4

−2

0

2

4

3 4 5 6 7 8 9 10 11


Unemployment rate



3 4 5 6 7 8 9 10 11

Unemployment rate





π t − E

π tBA

CK

PC

E in

flatio

n

π t − E

π tBA

CK

GD

P d

efla

tor

infla

tion

π t − E

π tBA

CK

CP

I CO

RE

infla

tion

π t − E

π tSP

F

π t − E

π tBA

CK

part

ial o

ut o

il pr

ice

chan

ges

π t − E

π tBA

CK

P

CE

CO

RE

infla

tion


Note: See notes for Figure 1 and the text for more details.Source: Coibion and Gorodnichenko (2015) – SPF forecast over next four quarters.


POSSIBLE EXPLANATIONS

Does unemployment mismeasure forcing variable?

Has the slope of the Phillips curve declined?

Are inflation expectations by firms mismeasured?


ACCOUNTING FOR NATURAL RATE

More natural to think of ut − unt as forcing variable

Perhaps the natural rate of unemployment changed

during Great Recession

Two approaches:

Use CBO estimates of natural rate of unemployment

Ask what natural rate consistent with observed inflation dynamics?



Reestimating Phillips curve with unemployment gap has

no effect on slope

CBO estimate of natural rate only rose by 1 percentage point

in Great Recession

Minimal effect on missing disinflation


ACCOUNTING FOR NATURAL RATEVOL.7 NO.1 207COIBION AND GORODNICHENKO: THE MISSING DISINFLATION

0

2

4

6

8

10

12

2007 2008 2009 2010 2011 2012 2013

07Q4

08Q1

08Q2

08Q3

09Q1

09Q2

09Q309Q4

10Q1

10Q2

10Q3

10Q4

11Q1

11Q2

11Q3

11Q4

12Q1

12Q2

12Q312Q4

13Q1

−8

−6

−4

−2

0

2

4

−2 −1 0 1 2 3 4 5

Unemployment gap

Panel A. Missing disin�ation with CBO unemployment gaps

Panel B. Changes in natural rate of unemployment needed to explain missing disin�ation

π t − E

π tBA

CK

Predicted natural rate

95% CI for predicted rate

Actual CBO natural rate

Actual unemployment rate

Figure 3. Does the Missing Disinflation Reflect Structural Unemployment?

Notes: Panel A plots quarterly levels of the unemployment gap (the difference between actual unemployment and the CBO estimate of the short-term natural rate of unemployment) against quarterly deviations of inflation from expected inflation (measuring the latter as the average inflation rate over the previous four quarters). The trend line uses data from 1960Q1 to 2007Q3. The predicted natural rate of unemployment in panel B is estimated as follows. First, U E t − U E t n = α + β ( π t − E t π t+1 ) + ε t is estimated on the 1960Q1–2007Q3 sample, where U E t is the rate of unemployment, U E t n is the natural rate of unemployment from the CBO, E t π t+1 is the backward-looking mea-sure of inflation expectations. Second, predicted value of the natural rate of unemployment is UE t

n = U E t − α −

β ( π t − E t π t+1 ) . The solid line shows the path of UE t n , while the shaded region shows the 95 percent confidence

interval for the predicted value. The solid line with circle markers is the natural rate of unemployment from the CBO. The dashed line shows the path of actual unemployment rate.




VOL.7 NO.1 207COIBION AND GORODNICHENKO: THE MISSING DISINFLATION

0

2

4

6

8

10

12

2007 2008 2009 2010 2011 2012 2013

07Q4

08Q1

08Q2

08Q3

09Q1

09Q2

09Q309Q4

10Q1

10Q2

10Q3

10Q4

11Q1

11Q2

11Q3

11Q4

12Q1

12Q2

12Q312Q4

13Q1

−8

−6

−4

−2

0

2

4

−2 −1 0 1 2 3 4 5

Unemployment gap

Panel A. Missing disin�ation with CBO unemployment gaps

Panel B. Changes in natural rate of unemployment needed to explain missing disin�ation

π t − E

π tBA

CK

Predicted natural rate

95% CI for predicted rate

Actual CBO natural rate

Actual unemployment rate

Figure 3. Does the Missing Disinflation Reflect Structural Unemployment?

Notes: Panel A plots quarterly levels of the unemployment gap (the difference between actual unemployment and the CBO estimate of the short-term natural rate of unemployment) against quarterly deviations of inflation from expected inflation (measuring the latter as the average inflation rate over the previous four quarters). The trend line uses data from 1960Q1 to 2007Q3. The predicted natural rate of unemployment in panel B is estimated as follows. First, U E t − U E t n = α + β ( π t − E t π t+1 ) + ε t is estimated on the 1960Q1–2007Q3 sample, where U E t is the rate of unemployment, U E t n is the natural rate of unemployment from the CBO, E t π t+1 is the backward-looking mea-sure of inflation expectations. Second, predicted value of the natural rate of unemployment is UE t

n = U E t − α −

β ( π t − E t π t+1 ) . The solid line shows the path of UE t n , while the shaded region shows the 95 percent confidence

interval for the predicted value. The solid line with circle markers is the natural rate of unemployment from the CBO. The dashed line shows the path of actual unemployment rate.



HAS THE PHILLIPS CURVE FLATTENED?212 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS JANUARY 2015

curve. Hence, it appears plausible that a flattening of the Phillips curve could explain some of the missing disinflation.

We investigate the statistical evidence for a change in the slope of the Phillips curve more formally by allowing for a break in the slope of the Phillips curve in 1985Q1 as follows:

(5) π t − E t π t+1 = c + κ × U E t gap + γ × U E t gap × I ≥85,t + θ × I ≥85,t + erro r t ,

07Q4

08Q108Q2

08Q3

09Q1

09Q2

09Q3

09Q410Q110Q2

10Q310Q411Q111Q2

11Q3

11Q4

12Q1

12Q2

12Q3

12Q413Q1

−2

0

2

4

−3 −1 0 1 3 4 5

Unemployment gap

1960 Q1−1984 Q4

1985 Q1−2007 Q3

2007 Q3−2013 Q1

−10

−8

−6

−4

−2

0

2

4

6

Infla

tion

(CP

I)

2008 2009 2010 2012 2013

07Q4

08Q1

08Q2

08Q3

09Q1

09Q2

09Q309Q4

10Q1

10Q2

10Q3

10Q4

11Q1

11Q2

11Q3

11Q4

12Q1

12Q2

12Q312Q4

13Q1

−6

−4

−2

0

2

4

−3 −2 0 1 2 3 4 5

Unemployment gap

−1

−2 2

2007 2011

1960 Q1−1984 Q4

1985 Q1−2007 Q3

2007 Q3−2013 Q1

Panel A. Sample split in mid-1980’s, backward-looking PC

Panel B. Sample split in mid-1980’s, forward-looking PC

Panel C. Counterfactual in�ation paths from time-varying slopes

Actual

BACK PC, 1960−2007

SPF PC, 1981−2007

BACK PC, 1991−2007

SPF PC, 1991−2007

π t−

Eπ t

SP

F ,

GD

P d

efla

tor

π t−

Eπ t

Bac

k

Figure 5. Time Variation in the Slope of the Phillips Curve

(Continued)



HAS THE PHILLIPS CURVE FLATTENED?


curve. Hence, it appears plausible that a flattening of the Phillips curve could explain some of the missing disinflation.

We investigate the statistical evidence for a change in the slope of the Phillips curve more formally by allowing for a break in the slope of the Phillips curve in 1985Q1 as follows:

(5) π t − E t π t+1 = c + κ × U E t gap + γ × U E t gap × I ≥85,t + θ × I ≥85,t + erro r t ,

07Q4

08Q108Q2

08Q3

09Q1

09Q2

09Q3

09Q410Q110Q2

10Q310Q411Q111Q2

11Q3

11Q4

12Q1

12Q2

12Q3

12Q413Q1

−2

0

2

4

−3 −1 0 1 3 4 5

Unemployment gap

1960 Q1−1984 Q4

1985 Q1−2007 Q3

2007 Q3−2013 Q1

−10

−8

−6

−4

−2

0

2

4

6

Infla

tion

(CP

I)

2008 2009 2010 2012 2013

07Q4

08Q1

08Q2

08Q3

09Q1

09Q2

09Q309Q4

10Q1

10Q2

10Q3

10Q4

11Q1

11Q2

11Q3

11Q4

12Q1

12Q2

12Q312Q4

13Q1

−6

−4

−2

0

2

4

−3 −2 0 1 2 3 4 5

Unemployment gap

−1

−2 2

2007 2011

1960 Q1−1984 Q4

1985 Q1−2007 Q3

2007 Q3−2013 Q1




Actual

BACK PC, 1960−2007

SPF PC, 1981−2007

BACK PC, 1991−2007

SPF PC, 1991−2007

π t−

Eπ t

SP

F ,

GD

P d

efla

tor

π t−

Eπ t

Bac

k

Figure 5. Time Variation in the Slope of the Phillips Curve

(Continued)



HAS THE PHILLIPS CURVE FLATTENED?

Phillips curve appears to have flattened after 1985

Although statistical significance is weak

(little variation in inflation after 1985)

Can this account for missing disinflation?


CAN THIS ACCOUNT FOR MISSING DISINFLATION?VOL.7 NO.1 213COIBION AND GORODNICHENKO: THE MISSING DISINFLATION

where I ≥85,t is a dummy variable equal to one for periods from 1985Q1 to 2007Q3 and zero prior to 1985. The interaction of this dummy variable with the unemploy-ment gap ( γ ) allows us to assess whether the slope of the Phillips curve changed around this period.

We report results from estimating this specification in Table 1 for several cases: CPI inflation and backward-looking expectations, GDP deflator inflation and back-ward-looking expectations, and GDP deflator inflation with forward-looking (SPF) expectations.7 In each case, we estimate the Phillips curve both by OLS and by IV, using as instruments a constant, one lag of unemployment, the dummy variable for post-84 periods, and the interaction of the dummy with the lag of unemployment.

The point estimates on the interaction term are always positive, so that the Phillips curve consistently appears to have flattened since the mid-1980s. However, the sta-tistical significance of this effect varies by specification: we cannot reject the null of no change in slope using the CPI but can reject the null at least at the 5 percent level for all specifications with the GDP deflator, with almost no difference between OLS and IV estimates in any case. Thus, the evidence for a change in the slope is mixed. What is consistent across specifications, however, is that the change in the slope (if there was one) was relatively large: on the order of a 60–80 percent reduction in most specifications. Furthermore, we cannot reject the null that the slope of the Phillips curve since 1985 was zero, pointing to a weak link between real and nominal economic activity during this period, consistent with Atkeson and Ohanian (2001). The absence of conclusive empirical evidence on a changing slope

7 We use GDP deflator inflation forecasts from the SPF because projections for the GNP/GDP deflator have been collected by the SPF since 1968 while CPI inflation projections have been collected by the SPF only since 1981. The short time series of the CPI projections in the SPF makes the analysis of structural breaks problematic.

07Q4

08Q108Q2

08Q3

09Q1

09Q2

09Q3

09Q410Q110Q2

10Q310Q411Q111Q2

11Q3

11Q4

12Q1

12Q2

12Q3

12Q413Q1

−2

0

2

4

−3 −1 0 1 3 4 5

Unemployment gap

1960 Q1−1984 Q4

1985 Q1−2007 Q3

2007 Q3−2013 Q1

−10

−8

−6

−4

−2

0

2

4

6

Infla

tion

(CP

I)

2008 2009 2010 2012 2013

07Q4

08Q1

08Q2

08Q3

09Q1

09Q2

09Q309Q4

10Q1

10Q2

10Q3

10Q4

11Q1

11Q2

11Q3

11Q4

12Q1

12Q2

12Q312Q4

13Q1

−6

−4

−2

0

2

4

−3 −2 0 1 2 3 4 5

Unemployment gap

−1

−2 2

2007 2011

1960 Q1−1984 Q4

1985 Q1−2007 Q3

2007 Q3−2013 Q1




Actual

BACK PC, 1960−2007

SPF PC, 1981−2007

BACK PC, 1991−2007

SPF PC, 1991−2007

π t−

Eπ t

SP

F ,

GD

P d

efla

tor

π t−

Eπ t

Bac

k

Figure 5. (Continued)

Notes: Panels A and B show changes in the slope of the Phillips curve over time. Panel A uses CPI inflation rate. Panel B uses GDP deflator inflation rate. Panel C shows the path of actual inflation and inflation predicted by Phillips curve estimated with forward-looking expectations (SPF) and backward-looking expectations (BACK) and over different time samples.



INFLATION EXPECTATIONS

Perhaps inflation expectations of firms differ from backward-looking

expectations and SPF forecasts

Household expectations (Michigan survey) quite different


INFLATION EXPECTATIONS218 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS JANUARY 2015

well-proxied by professional forecasts. While one might expect very large firms to have professional forecasters on staff or to rely on the services of professional forecasters to guide their economic decisions, this need not be the case for small and medium enterprises for whom the gains from having precise information about aggregate conditions may be small (especially relative to local or industry-specific conditions), as in Mackowiak and Wiederholt (2009). For such firms, household forecasts could very well be a better proxy of their beliefs than professional forecasts.

Does it matter for the Phillips curve and the missing disinflation whether one assumes that firms hold beliefs closer to those of professional forecasters or house-holds? We showed in panel E of Figure 1 that using professional forecasts of inflation did not meaningfully affect the estimated slope of the historical Phillips curve or the presence of missing disinflation during the Great Recession. In panel B of Figure 6, we present the Phillips curve relationship between the unemployment gap and the difference between CPI inflation and household expectations of inflation. Several

Infla

tion

Unemployment gap

−10

−8

−6−4−2

02

46

8

Infla

tion

2007

2008

2009

2010

2011

2012

2013

−10

−8−6−4−2

02468

2011

2013

07Q408Q1

08Q2

08Q3

09Q1

09Q2

09Q309Q4

10Q1

10Q2

10Q3

10Q411Q111Q211Q3

11Q412Q212Q312Q4

13Q1

−6

−4

−2

0

2

4

−3 −2 −1 0 1 2 3 4 5

0

2

4

6

8

10

12

1980 1985 1990 1995 2000 2005 2010

2012

2007

2008

2009

2010

ActualBackward PCSPF PCMSC PC

Panel A. In�ation expectations for different economic agent

Panel B. Phillips Curve with household in�ation expectations

Panel C. Counterfactual in�ation (CPI) paths for different expectations

Panel D. Relative contribution of slopes and in�ation expectations

1960Q1−1984Q41985Q1−2007Q32007Q3−2013Q1

MSC exp, SPF κMSC PC MSCκ, SPF expSPF PC

Actual

Asset pricesMichiganSPF (CPI)

π t −

Eπ t

MSC

12Q1

Figure 6. The Phillips Curve and the Missing Disinflation with Household Inflation Expectations

Notes: Panel B shows the scatter plot of inflation (CPI) surprises versus unemployment gap as well as fitted linear regressions for two subperiods. Panel C plots actual inflation rate (CPI) as well as inflation rate predicted by Phillips curves (equation (1)) estimated on the pre-Great Recession samples. Phillips curves are estimated with unemploy-ment gap as the forcing variable. Panel D presents decomposition of differences between predicted inflation rates from Phillips curves (equation (1)) estimated with inflation expectations from the Michigan Survey of Consumers (MSC) and Survey of Professional Forecasters (SPF). “XYZ exp” denotes which inflation expectations are used (SPF or MSC), while “XYZ κ ” denotes what data was used to estimate the slope of the Phillips curve. Phillips curves are estimated with unemployment gap as the forcing variable.




Not clear whether firm expectations behave like SPF

or like household expectations

But if they do behave like household expectations, does it matter?


INFLATION EXPECTATIONS218 AMERICAN ECONOMIC JOURNAL: MACROECONOMICS JANUARY 2015



Infla

tion

Unemployment gap

−10

−8

−6−4−2

02

46

8

Infla

tion

2007

2008

2009

2010

2011

2012

2013

−10

−8−6−4−2

02468

2011

2013

07Q408Q1

08Q2

08Q3

09Q1

09Q2

09Q309Q4

10Q1

10Q2

10Q3

10Q411Q111Q211Q3

11Q412Q212Q312Q4

13Q1

−6

−4

−2

0

2

4

−3 −2 −1 0 1 2 3 4 5

0

2

4

6

8

10

12

1980 1985 1990 1995 2000 2005 2010

2012

2007

2008

2009

2010






1960Q1−1984Q41985Q1−2007Q32007Q3−2013Q1


Actual


π t −

Eπ t

MSC

12Q1






Three differences versus SPF:

No evidence of flattening

Flatter throughout

No evidence of missing disinflation!






Infla

tion

Unemployment gap

−10

−8

−6−4−2

02

46

8

Infla

tion

2007

2008

2009

2010

2011

2012

2013

−10

−8−6−4−2

02468

2011

2013

07Q408Q1

08Q2

08Q3

09Q1

09Q2

09Q309Q4

10Q1

10Q2

10Q3

10Q411Q111Q211Q3

11Q412Q212Q312Q4

13Q1

−6

−4

−2

0

2

4

−3 −2 −1 0 1 2 3 4 5

0

2

4

6

8

10

12

1980 1985 1990 1995 2000 2005 2010

2012

2007

2008

2009

2010






1960Q1−1984Q41985Q1−2007Q32007Q3−2013Q1


Actual


π t −

Eπ t

MSC

12Q1





IMPORTANCE OF VOLCKER DISINFLATION

Much of evidence for slope on Phillips curve comes from

Volcker Disinflation

Details turn out to matter for results

Important methodological change in CPI in 1983

(housing switched to rental equivalence)

12-month inflation versus quarterly inflation matters

No missing disinflation with 12-month PCE


MISSING DISINFLATION? +

Pre‐2008 Post‐2008 Missing Disinflation?

4‐quarter PCE inflation, SPF exp.

‐0.09(0.05)

‐0.06(0.06)

No

4‐quarter core PCE,SPF exp.

0.27(0.03)

‐0.08(0.05)

No

Quarterly PCE, SPF exp.

‐0.29(0.07)

0.02(0.12)

Yes

Quarterly PCE, 4‐quarter lagged exp.

‐0.31(0.05)

0.10(0.16)

Yes

Slope Coefficients () from Different Specifications

Source: Nakamura (2018): Discussion at AEA Meetings


MISSING DISINFLATION?



DID Etπt+1 DO ALL THE WORK?



DID Etπt+1 DO ALL THE WORK?

Source: Nakamura (2018): Discussion at AEA Meetings – Quarterly PCE rather than 12-month


Empirical Assessment of the Phillips Curve · 2018. 2. 5. · Slope of Phillips curve: Causal...

Documents

Transcript of Empirical Assessment of the Phillips Curve · 2018. 2. 5. · Slope of Phillips curve: Causal...