Transcript Slide 1

The Phillips Curve & US Monetary Policy:
How Does the FOMC Forecast Inflation?
Ellen Meade
American University
Dan Thornton
Federal Reserve Bank of St. Louis
Objective of analysis
• Examine use of Phillips curve concept in US monetary
policy decisionmaking
• 3 aspects of analysis:
• Search FOMC transcripts using key words
•
•
•
•
1982 through 2002
key words: potential, gap, Phillips curve, NAIRU
identify portions of text to read
assess use of Phillips curve
• Analyze Fed staff (Greenbook) inflation forecasts
• how accurate are they
• do they appear to have been generated using a NK Phillips curve
• sacrifice ratio implied by forecast
• Was Phillips curve concept important for setting Federal funds
target?
Literature
• Vast literature on Phillips curve
• Reference 5 main things:
(1) Goodfriend & King (2005) study on Volcker
disinflation and inflation expectations
(2) Orphanides work on real-time data
(3) papers that forecast inflation using simple Phillips
curve models, such as Atkeson & Ohanian (2001), Ball &
Mankiw (2002), and Staiger, Stock & Watson (1997)
(4) Bernanke speech on inflation -- how the Fed staff
puts together the Greenbook inflation projections
(5) King (2008) paper on Phillips curve and US economic
policy, 1958-1996
Methodology
• Search transcripts using key words
– read surrounding discussion
• 800 interventions in 168 transcripts
– “intervention” is separate remark
– key word may have been used multiple times in single
intervention
– 2 key words used in same intervention counts twice
• Large increase in “hits” for all key words starting in 1994
for policymakers and staff
References to key words by Federal Reserve policymakers and staff in FOMC transcripts,
1982-2002*
Year
“Potential”
Output
“Phillips”
“NAIRU”
Staff
output
“gap”
curve
Total
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
0
1
2
2
1
3
9
13
5
1
9
6
39
42
45
26
26
31
52
35
29
0
0
0
0
0
2
1
2
1
1
2
2
3
9
7
13
1
16
9
9
17
1
3
0
0
0
2
2
5
4
0
0
1
10
4
12
8
2
9
3
0
6
0
0
0
0
0
0
1
0
0
0
0
1
26
29
36
32
11
35
54
6
12
0
0
1
0
0
2
6
2
5
0
4
3
15
18
3
29
23
35
65
51
37
Total
377
95
72
243
299
* Each key word is counted once per policymaker intervention, even if the key word is used multiple
times during that intervention. Entries are computed using all available transcripts of meetings and
conference calls in a given year, exclude staff interventions, and include Donald Kohn from August
2002 when he moved from the staff to become a Federal Reserve official.
FOMC in 1980s
• Limited reference to concept, usually comes up when
Fed staff is explaining Greenbook inflation forecast (first
part of meeting)
• Not referred to as “Phillips curve” or with reference to
NAIRU, but in terms of
– available slack in the economy
– level of operating capacity
– some measure of resource utilization
• Clear in exchanges with the staff, however, that they
have based inflation forecast on a Phillips curve model
• No real challenges to framework by policymakers
• From arrival in late 1980s, Greenspan indicates
skepticism
Typical exchange
• Fed Governor Martha Seger and staff member Michael
Prell (9/1988, p. 5)
Seger: “… My second question is one involving a statement in
the Greenbook that staff continue to believe that additional
pressure in financial markets would be required to slow the
expansion… Do you think you’ve seen the full impact of the
tightening moves that we’ve had to date?”
Prell: “No, but our thought is that even after we have absorbed
those, that we will not open up enough slack, so to speak, or
reduce the pressures on the economy enough, to relieve the
inflationary pressures. Therefore, we need to hold growth below
potential for a period of time. What we have in our forecast is a
very slight shortfall of actual growth from potential and very slight
easing pressures on resources. And we think that…, we are
going to need a bit more restraint to keep things under control…”
Greenspan the skeptic
• “When we look at capacity—and the Fed is the official
source of these data—capacity is a very dubious
concept. You really don’t know whether or not you have
run into capacity until you have some objective
measures of the inability to meet customer orders. That
is really what it’s all about.” (3/1988, pp. 40-41)
• “That depends to a large extent on the accuracy of the
Phillips curve type of model.” (2/1990, p. 26)
"Potential" output
Part 1
Part 2
40
35
30
25
20
15
10
5
2002
1997
1992
1987
1982
0
"Gap"
20
18
16
14
12
10
8
6
4
2
2002
1997
1992
1987
1982
0
"NAIRU"
45
40
35
30
25
20
15
10
5
20
02
19
92
19
82
0
"Phillips Curve"
12
10
8
6
4
2
2002
1997
1992
1987
1982
0
Roadmap, 1994 and beyond
• Increase in academics on Fed Board
– Blinder, Yellen, Meyer, Gramlich
• Increase in overt discussion of Phillips curve, Taylor rule,
and other economic models or concepts
• By end of 1994, economy at or above full employment
• Fed staff’s forecast for inflation begins to go off track in
mid-1994
–
–
–
–
over predicts inflation until end-1998, with rising errors over time
particularly large errors from February 1997
misses improvement in productivity
compensate in forecast by reducing estimate of NAIRU
Interventions by Blinder, Meyer, and Yellen
(percent of total interventions)
1994
1995
1996
1997
1998
1999
2000
“Potential”
output
Output
“gap”
“Phillips”
curve
“NAIRU”
11.9
20.9
23.8
3.7
7.7
0.0
20.8
33.3
88.9
28.6
7.7
0.0
12.5
11.1
20.0
50.0
16.7
12.5
50.0
11.1
66.7
19.2
20.7
50.0
56.3
0.0
14.3
20.0
-2
-1
0
1
2
Errors in Greenbook forecasts for consumer price inflation
(actual minus predicted), 1988-2000*
01 Jan 88
01 Jan 92
01 Jan 90
01 Jan 96
01 Jan 00
01 Jan 94
01 Jan 98
FOMC meeting date
* Forecast is for inflation in the calendar year following the year of the
FOMC meeting. Actual data are historical data as of 2007:Q1.
A debate emerges
• Policymakers divide into 2 camps: defenders of the
Phillips curve and critics of it
– defenders use Phillips curve model and forgive errors
• focus on estimate of NAIRU and whether to change it
• how sensitive is inflation forecast to alternative NAIRU estimates
– critics want alternative models and worry that inflation may rise
unexpectedly, argue for policy tightening
• Cleveland Fed President Jerry Jordan asks staff for alternative
frameworks for forecasting inflation (5/1996)
• Greenspan remains on the fence, very eclectic and
always a skeptic
– “... Saying that the NAIRU has fallen, which is what we tend to
do, is not very helpful. That’s because whenever we miss the
inflation forecast, we say the NAIRU fell.” (12/1995, p. 39)
Wrap up of transcripts
• Much arguing about correct estimate of NAIRU and/or
whether Phillips curve model should be used to forecast
inflation (number of policymakers who disregard Phillips
curve model rises over time)
• More discussion at June 1999 FOMC meeting triggered
by presentation of Orphanides work on real-time output
gaps
• By end of 1990s, many FOMC members find Phillips
curve concept useless and are searching for other ways
to forecast inflation
• June 2002 FOMC meeting – presentation of “Are Phillips
Curves Useful for Forecasting Inflation? 40 Years of
Debate”
Empirical analysis
• Examine Greenbook inflation forecasts to see how much
they seem to rely on Phillips curve model
• Explore how much variation in projected inflation or
projected change in inflation explained by output or
unemployment gap
– Greenbook inflation in current and next calendar year
• focus on core inflation
– Greenbook output gaps 1987 to 2000 (current quarter, lag
quarter, lead quarter), 112 meetings
– Greenbook unemployment rate current and next calendar year;
CBO real-time NAIRU from 1991
– look at very simple Phillips curve equations
Bernanke on Greenbook forecasts (1)
• Bernanke speech “Inflation Expectations and Inflation
Forecasting”, NBER July 2007
• Fed staff Greenbook long-term track record in
forecasting inflation is quite good by any reasonable
benchmark (Romer & Romer, 2000; Faust & Wright,
2007)
• Staff employs variety of formal models, both structural
and statistical, in its forecasting efforts
• Forecasts for the very near term—built from bottom up
• Forecasts beyond two quarters—built from top down
Bernanke on Greenbook forecasts (2)
• … forecasts of inflation… developed through an eclectic
process that combines model-based projections,
anecdotal and other ‘extra-model’ information, and
professional judgment
• Forecasting horizons beyond a quarter or two
– Inflation´s fundamental determinants
– Food and energy forecasted separately from the core
– Most of the models used are based on versions of the New
Keynesian Phillips curve, which links inflation to inflation
expectations, the extent of economic slack, and indicators of
supply shocks
• Measuring inflation expectations
– Popular shortcut is to include lagged inflation terms
– Drawbacks lead to use of survey measures
Estimation
• 2 simple Phillips curve equations:
corey+1 = α + β1corey + β2y-gapq–1 + β3y-gapq + β4y-gapq+1
corey+1 = α + β1corey + β2u-gapy + β3u-gapy+1
• Estimated using OLS (robust standard errors)
– output gap terms improve fit in equation estimated over entire
period
• Examine fit (adjusted R-squared) from rolling
regressions, 50-meeting window (6+ years)
Chart 3. Adjusted R-squared from rolling regressions (50-meeting window) of Phillips
curve equations using output gaps
0
.2
.4
.6
.8
1
Adjusted R-Squared
01 Jan 93
01 Jul 94
01 Jan 96
01 Jul 97
FOMC meeting date
Inflation only
Marginal contribution from Output gaps
01 Jan 99
01 Jul 00
Inflation + Output gaps
Chart 4. Adjusted R-squared using unemployment gaps
0
.2
.4
.6
.8
1
Adjusted R-Squared
01 Jan 93
01 Jul 94
01 Jan 96
01 Jul 97
FOMC meeting date
01 Jan 99
01 Jul 00
Inflation only
Inflation + Unemployment gaps
Marginal contribution from Unemployment gaps
Sacrifice ratio
• Computed implicit sacrifice ratio as current-year
unemployment gap divided by projected one-yearahead change in core inflation
• Generally, sacrifice ratio is negative
– not always so when forecast for total inflation used
– very small/negligible in mid-1990s
– December 1997 to December 1998, large and positive
-20
-10
0
Sacrifice ratio
10
20
Chart 5. Sacrifice ratio (unemployment gap in calendar year of FOMC meeting divided
by projected calendar-year change in core CPI inflation)
01 Jan 91
01 Jul 93
01 Jan 96
FOMC meeting date
01 Jul 98
01 Jan 01
Conclusions
• To what extent were policy decisions based upon Phillips
curve concept?
• Very difficult to know
• Greenspan was eclectic, not wedded to Phillips curve
– always put forth proposal which was adopted by FOMC
– could this mean that Phillips curve does not underlie decisions?
• Target for Federal funds rate and Greenbook inflation
forecast suggest a period (August 1996 to December
1997) when funds target virtually unchanged but inflation
forecast fell steeply
4
2
3
6
4
2
Fed funds target
8
Greenbook CPI inflation yr+1
5
10
Chart 6. The Federal funds target (left scale) and the Greenbook forecast for CPI
inflation in the calendar year following the FOMC meeting (right scale)
01 Jan 86
01 Jan 90
01 Jan 94
01 Jan 98
01 Jan 88
01 Jan 92
01 Jan 96
01 Jan 00
FOMC meeting date
Fed funds target
Greenbook CPI inflation yr+1