BRUSSELS ECONOMIC REVIEW - CAHIERS ECONOMIQUES DE BRUXELLES

VOL. 53 (3/4) AUTUMN - WINTER 2010

THE NEW KEYNESIAN PHILLIPS CURVE

FOR THE EURO AREA: WHICH SPECIFICATION?

EMMANUEL CARRÉ

ABSTRACT:
The New Keynesian Phillips Curve (NKPC) specification in the Euro area is controversial.
For some it is purely forward-looking, when others incorporate a backward-looking
component into a ‘hybrid’ NKPC (HNKPC). This article contributes to this debate first by
drawing stylised facts of the Euro area (H)NKPC from a quantitative review of 88
comparable empirical estimates from 21 articles. If the forward-looking share dominates, a
substantial backward-looking fraction is found (about one-third). This result is fairly robust
to sub-sample and simple tests. It is also shown that micro evidence from firms’ pricing
surveys is favourable to this result. Since the observed variations in estimates could be a
limitation to this otherwise clear result, the second contribution of this article is to explore the
factors driving these variations. Finally, policy implications are examined.

JEL CLASSIFICATION : E31; E52; E58.

KEYWORDS: New Keynesian Phillips Curve, Inflation Persistence.


University Paris XIII - CEPN. Correspondence address: carre.emnl@gmail.com. I thank referees
for useful suggestions, Edwin Le Héron and Alexander Mihailov for helpful comments and Isabelle
Salle for very important discussions. Any remaining errors are my own.

337
 THE NEW KEYNESIAN PHILLIPS CURVE FOR THE EURO AREA:
WHICH SPECIFICATION?

INTRODUCTION

The New Keynesian Phillips Curve (NKPC), new version of the old curve
proposed by Phillips (1958), is popular in the academic literature and is ‘the
closest thing there is to a standard specification’ (McCallum, 1997:357). It is
the standard inflation theory of the ‘New Consensus’ in macroeconomics: the
New Neo-Classical Synthesis (NNS)1. It is also the core equation of the ‘state of
the art’ in business cycles modelling, the Dynamic Stochastic General
Equilibrium (DSGE) models2. It is micro-founded on the New Keynesian theory
of price stickiness cum rational expectations hypothesis (Calvo, 1983). This
results in a purely forward-looking formula as the baseline or ‘canonical’
NKPC3 describing inflation as a purely ‘expectational’ phenomenon, a ‘jump’
and not a persistent variable. With these micro-foundations it should surpass the
old Phillips curve or its expectations augmented and accelerationist versions that
had the disadvantages of emerging from empirical results and being mostly
backward-looking, with little emphasis on expectations4. Furthermore, it is
supposed to be an empirically valid description of inflation dynamics, in the
euro area in particular (Gali et al., 2001:1237). At the centre of the New
Keynesian ‘science’ of optimal monetary policy, the NKPC recommends
policymakers to pursue an ‘expectation management strategy’ relying on
communications5. Indeed it asserts that if the central bank controls future
inflation expectations, it controls the present value of inflation.

These advices from the NKPC are apparently accepted by central bankers,
particularly the European Central Bank (ECB) that endorses the DSGE class of
models with Calvo pricing6. The ECB also integrates the emphasis on
expectations in monetary policy transmission mechanisms, the expectations
management strategy as well as the communication policy7. Hence, the NKPC
specification is a vital policy question.

Nevertheless, beyond the veil of polite academic and central bankers’

consensus, the Calvo-NKPC specification is controversial: ‘there are reasons,
however, to be somewhat dissatisfied with this state of affairs’ (McCallum,
2008:1481). Empirically, the purely forward-looking NKPC seems unable to
reproduce the inflation dynamics observed in euro area data8; in particular
inflation persistence, i.e. the hump-shaped, gradual and delayed response of
inflation to a shock9. This empirical failure is a severe drawback because
inflation persistence is considered as a stylised fact of European inflation

1
Goodfriend (2007).
2
A benchmark DSGE model is Smets and Wouters (2003).
3
It is generally considered that the analysis of the NKPC began with Roberts (1995).
4
See Gali and Gertler (2007:33): ‘Another important way that the new Phillips curve differs from
the old is that it is fully forward-looking’.
5
Woodford (2003 :15).
6
On its web site, the ECB mentions the Smets and Wouters model as one of the ‘macroeconomic
models of the euro area’. See http://www.ecb.int/home/html/researcher.en.html.
7
For the use of the expectations management, see ECB, Monthly Bulletin, May 2009, p.86.
8
As acknowledged by NKPC promoters themselves: ‘The macroeconomic variables within the
baseline model appear to display greater persistence in practice than the basic framework can
capture’ (Gali and Gertler, 2007:42).
9
See Mankiw (2001).

338
 EMMANUEL CARRÉ

dynamics by some New Keynesians10 and is a research subject of the ECB11. It

also suggests that the purely forward-looking NKPC is not a perfectly robust
theory of inflation that can be used in practice by central bankers without any
limitations.

Due to these limits, the purely forward-looking NKPC is increasingly

challenged by a hybrid NKPC (HNKPC). This includes a backward-looking
component to deal with the empirical question of inflation persistence, and more
generally intends to reconcile the theory of inflation with empirical inflation
dynamics. One thus returns, to a certain extent, to the previous accelerationist
Phillips Curve which relied on adaptive expectations.

This NKPC versus HNKPC debate is magnified by a controversy on the optimal

econometric methodology to measure the respective forward-looking and
backward-looking fractions. Different econometric methods of estimation give
conflicting results since it could be empirically mainly either forward or
backward-looking: there is ‘an extreme range of opinions from researchers
examining the same equation’ (Rudd and Whelan, 2007:167).

This article makes two contributions to this debate on euro area (H)NKPC
specification. First, in the spirit of the meta-analysis technique, a quantitative
review of empirical estimates is realised. It results in stylized facts on the
empirical specification of the (H)NKPC in the euro area. This method of
evaluation of the forward and backward fractions of the HNKPC seems
appropriate in the face of the intractable controversy on the empirical estimates
and estimation techniques of the NKPC. Results of this quantitative analysis
could indicate a basic ‘calibration’ of the euro area HNKPC. The second
contribution to the euro area HNKPC discussion is the analysis of factors
driving heterogeneity in reported empirical estimates. Indeed, the observed
variations in estimates could be a limitation to the otherwise clear result of the
quantitative review.

The article is organised as follows. The first section explains the competing
specifications of the HNKPC and the quantitative review methodology chosen
in the article. The second section analyses results, and checks their robustness
by inspecting HNKPC empirical microfondations, i.e. the consistency of macro
evidence with micro evidence from surveys of firms’ prices formation. This
second section also investigates the factors explaining the observed variations in
estimates using a regression analysis. Finally, the third section draws
implications of the results for the conduct of monetary policy in the euro area.

10
See O’Reilly and Whelan (2004), Angeloni et al. (2006) for evidence on inflation persistence in
the Euro area.
11
The ECB has an ‘Inflation Persistence Network’:
http://www.ecb.int/home/html/researcher_ipn.en.html

339
 THE NEW KEYNESIAN PHILLIPS CURVE FOR THE EURO AREA:
WHICH SPECIFICATION?

1. METHODOLOGY

A quantitative survey of empirical estimates of the HNKPC in the euro area is

performed, aimed at quantifying the forward and backward-looking shares. We
extend the HNKPC estimates review realised by Altissimo et al. (2006:17) and
Henzel and Wollmershäuser (2008:827) to build an estimates database. The
survey includes 88 empirical estimates of the HNKPC from 21 articles, for the
period from 2001 to 2010. To our knowledge, it is the first time such a vast
quantitative review is carried out12. The econometric analysis of the database
follows the quantitative review methodology employed in De Grauwe and Costa
Storti (2004). We first give the reasons for the choice of a quantitative survey,
then the data selection process is presented.

1.1. WHY A QUANTITATIVE SURVEY?

The (H)NKPC debate: three competing specifications. The first reason for the
quantitative survey choice is conflicting estimation results: the NKPC could be
empirically mainly either forward or backward-looking (Jondeau and Le Bihan,
2005:522). There are three competing views regarding the (H)NKPC empirical
specification that can be presented on the basis of its more general form, the
HNKPC, that includes both a forward-looking share (βf) and a backward-
looking share (βb):

 t   f E t  t 1   b  t 1  kx t (1)

where π is the inflation rate, x the output gap with x = y-y*, y represents the
output and y* the potential output, finally E is the expectation operator (rational
expectations hypothesis).
The mainstream view on the (H)NKPC (specification 1) defends a purely
forward-looking form (βf=1; βb=0)13:

 t   E t  t 1  kx t (2)

They claim that βb is different from 0, but is small enough to be neglected: ‘the
estimates imply that backward-looking price setting […] has been a relatively
unimportant factor behind the dynamics of Euro area inflation’ (Gali et al.,
2001:1256). On the contrary, advocates of specification 2 consider that βb>0,
but admit that the forward-looking share dominates (βf>βb). Finally,
specification 3 defenders state that the HNKPC is predominantly backward-
looking (βf<βb)14. This view is relatively similar to the traditional accelerationist
Phillips curve. Controversy on (H)NKPC estimation technique is the second
reason for the choice of a quantitative review.

12
As a comparison Altissimo et al. (2006:17) and Henzel and Wollmershaüser (2008) review
respectively 6 studies (11 estimates) and 6 studies (12 estimates) of the euro area HNKPC.
13
See Hondroyiannis et al. (2008).
14
See Fuhrer (1997), Rudd and Whelan (2005), Lindé (2005), Henzel and Wollmershaeuser (2008).

340
 EMMANUEL CARRÉ

An intractable controversy on the NKPC estimation technique. This debate on

the (H)NKPC empirical specification could partly result from econometric
techniques, with the βb value varying considerably depending on the estimation
method. The traditional method to estimate the HNKPC, the Generalised
Method of Moments (GMM), has been criticised for its bias (overestimating βf)
and its weak identification problem leading to a lack of reliability of its
estimates15. This is why a Full Information Maximum Likelihood (FIML)
method is often preferred16, but there are some indications it overestimates βb17.
Hence, some authors consider it can generate fragile results, so they prefer a
Bayesian estimation18. Accordingly, there is an intractable controversy on the
econometric technique to empirically evaluate the HNKPC: ‘the debate on the
appropriate technique to use when estimating Phillips curve remains open’
(Barkbu et al., 2005:6). This dispute is the second reason for the choice of a
quantitative review, having the advantage of offering results that can be judged
more consensual than those coming from a single article or estimation technique
considered in isolation.

Numerous other arguments justify this quantitative approach. First, several

studies doing this exercise have been mentioned, testifying it is common in the
literature. Secondly, it permits the combination, integration and comparison of
disparate estimates coming from various techniques, thus allowing for their
global comparative assessment, facilitating a comprehensive survey of the
literature. Thirdly, it suits to the growing volume of literature on the empirical
NKPC since, due to the large number of studies on this issue, a narrative survey
and synthesis would only have been manageable with difficulty. It is the
appropriate technique to review this flourishing literature while preserving a
synthetic view of its current state. Nevertheless, this review should be done
carefully, requiring the construction a robust database, based on rigorous criteria
of exclusion and inclusion that are now exposed.

1.2. DATA SELECTION PROCEDURES

Collected estimates concern the coefficients βf and βb (eq. 1). Estimates for the
HNKPC are selected, and not for the NKPC, because the latter does not allow
for backward-looking elements, not permitting to investigate the debate between
the three competing specifications. Inspired from the meta-analysis approach,
the selection of estimations available in the literature has been made according
to the two criterions of coverage and precision. Homogeneity of estimates is the
third selection criteria19.

In terms of coverage, the papers chosen cover the period from 2001 and 2010
since to our knowledge the euro area HNKPC literature begins with Gali et al.
(2001). As to precision –that is to say if estimations are representative of the
literature – articles and working papers in EconLit, Repec-Ideas and central

15
See Ma (2002), Mavroeidis (2005), Benati (2008 :1026), Kleibergen et Mavroeidis (2009:293).
16
See Fuhrer (1997), Lindé (2005) or Jondeau and Le Bihan (2008).
17
Jondeau and Le Bihan (2005 :539).
18
Benati (2008 :1026).
19
I wish to thank the referee for stressing the necessity to elaborate more on this point.

341
 THE NEW KEYNESIAN PHILLIPS CURVE FOR THE EURO AREA:
WHICH SPECIFICATION?

banks websites were consulted. Journal articles is an incomplete resource

limiting the number of available estimations because the (H)NKPC is a recent
literature with many working papers still unpublished due to the publication
delays. Statistically non significant estimates are also included. Estimates
homogeneity is a major concern to make them comparable, and is set up via
several strict criteria. First, sample homogeneity is checked: articles with
relatively similar long sample of 20 or 30 years have been collected. Estimates
from comparatively shorter subsample (ten years) or very long sample (since
1945) are not included. Yet samples cannot be perfectly similar, leading
potentially to differences in estimates (an issue inspected later). Secondly, all
studies use quarterly data, inflation being always measured as GDP deflator,
usually from the data set of the ECB Area-Wide Model (Fagan et al., 2001).
Studies using annual data or other inflation measures have not been considered.
Moreover, all studies consider the euro area as a single economy. Barkbu et al.
(2005) is the only exception in the database, estimating the HNKPC for the
whole euro area but also for its approximation by the aggregation of France,
Germany and Italy. This procedure is accepted in the literature because these
three countries cover 70% of the area-wide GDP and are the three largest
countries in terms of population20. On top of that it serves as a robustness check
for the whole euro area estimates. The third homogeneity basis is the micro-
foundation: studies with micro-foundations for the HNKPC included in the
analysis are based only on the Calvo (1983) price setting, as in the reference
article of Gali et al. (2001). Other micro-foundations such as the Taylor (1980)
price setting, state dependent pricing or sticky information are not considered.
Furthermore, HNKPC based on inflation expectations from surveys instead of
rational expectations are not taken into account. The fourth criterion to respect
homogeneity is an exclusive focus on closed-economy estimates. Estimates with
an explicit open economy component in the HNKPC specification have been
excluded, even those incorporating implicitly open economy elements, for
instance via the marginal cost formulation. The final homogeneity factor is that
all estimates result from a general specification common across studies: the
HNKPC. It is preferred to the purely forward looking NKPC that is often, but
not always, rejected by the data for the euro area21, so that data sometimes
support the HNKPC. Studies estimating a HNKPC specification adding an extra
right hand side money term, are also not included. Due to all these selection
criteria, the quantitative does not pretend to be exhaustive.

Promoters of the microfounded HNKPC claim it rests upon structural

parameters thus is supposed to be relatively robust to the Lucas critique: its
parameters βf and βb are supposed to be time invariant, as confirmed by some
empirical results for the euro area22. Even if some studies stress this stability is
imperfect23, it is taken as a consensual result in the first step of the analysis. A
particular attention is given in the selection to estimates that fulfil βf+βb=1. This
condition implies that the HNKPC respects the natural rate hypothesis24.

20
As argued by Jondeau and Sahuc (2008:40) who find results for these three countries which ‘look
very similar’ to those for the whole euro area.
21
Gali et al. (2001:1249).
22
Jondeau and Le Bihan (2005 :542), Groen and Mumtaz (2008 :31).
23
Barkbu et al. (2005:25), Benati (2008 :1035).
24
See McCallum (2008:1487).

342
 EMMANUEL CARRÉ

Having considered all these factors, the database includes 88 comparable

estimates from 21 papers for the euro area (Table 1)25. The same weight is put
on each of these estimates despite the risk that results can be biased by the large
number of estimates from a single article that would be over-represented. The
solution of selecting the ‘best’ or more representative estimates is not chosen
because of the multiple selection criteria possible in practice. It risks introducing
subjective judgment and a potential selection bias. Results are now presented
and analysed.

TABLE 1. REFERENCES OF ESTIMATIONS INCLUDED IN THE QUANTITATIVE

REVIEW, CLASSIFIED BY DATE
Number of Estimation Driving variable:
References Sample
estimates techniques output gap
Gali et al. (2001)* 4 1970:1-1998:2 GMM No
Gagnon and Khan (2001) 9 1970:1-1998:4 GMM No; Yes
Bardsen et al. (2002) 4 1972:2-1998:1 GMM; TSLS No
Gali et al. (2003) 2 1970:1-1998:2 GMM No
McAdam and Willman (2003) 8 1970:1-1997:4 GMM No
Smets and Wouters (2003) 1 1980:2-1999:4 Bayesian No
Smets and Wouters (2004) 1 1980:1-2002:4 Bayesian No
Bardsen et al. (2004) 4 1971:3-1998:1 GMM; TSLS No
De Walque et al. (2004) 2 1970:1-1998:2 Bayesian No
Gagnon and Khan (2005) 15 1970:1-1998:4 GMM No
Barkbu et al. (2005) 6 1975:1-2002:4
GMM No
1971:2-2002:3
Fanelli (2005) 2 1971:1-1998:2 ML No; Yes
Jondeau and Le Bihan (2005) 12 1970:1-1999:4 GMM; ML No; Yes
Rumler (2005) 2 1970:1-1998:4 GMM No
De Walque, Smets and Wouters (2006a) 6 1970:1-2002:2 Bayesian No
De Walque, Smets and Wouters (2006b) 1 1970:2-2002:2 Bayesian No
Paustian and Pytlarczyk (2006) 3 1970:1-2003:4 Bayesian No
Lippi and Neri (2007) 2 1981:1-2004:4 ML; GMM No
Sahuc and Smets (2008) 1 1985:1-2004:4 Bayesian No
Hondroyiannis et al. (2008) 2 1970:1-2005:3 GMM; TVC No
Onatski and Williams (2010) 1 1970-1999 Bayesian No

Notes: GMM: Generalized Method of Moments; ML: Maximum Likelihood; TSLS: Two-Stage
Least Squares; TVC: time-varying coefficient. *: to take into account errors in this article, stressed
by Gali et al. (2003), only 4 estimates are considered and 2 are rejected. The latter are replaced by
those presented in Gali et al. (2003). ‘No’, ‘Yes’ in the last column of the table means that the study
is estimating a HNKPC without and with the output gap as the driving variable.

2. RESULTS

This quantitative review is interested in the NKPC versus HNKPC debate for
the euro area: that is to say the validity of the three competing specifications.
Results are examined in four steps. First, results are represented graphically and
descriptive statistics are studied. The same is done for results of the sub-sample
made of estimates respecting the natural rate hypothesis. Secondly, simple t-
tests are undertaken to give indication of the strength of the three competing
specifications. Thirdly, results are confronted to the previous literature and
micro-level evidence. Finally, a regression analysis is undertaken to analyse
results further.

25
Database available upon request.

343
 THE NEW KEYNESIAN PHILLIPS CURVE FOR THE EURO AREA:
WHICH SPECIFICATION?

2.1. DESCRIPTIVE STATISTICS

All estimations. Quantitative review results are reported in the scatter graph
(Figure 1). The visual impression is one where the βb component is not nil.
Consequently, specification 1, the purely forward-looking NKPC, does not seem
observed since only a minority of estimates is on the horizontal axis. On the
contrary, a majority of estimates favours the HNKPC. However, most of them
are in the scatter graph bottom right-hand corner, visually indicating that βf
dominates βb. It gives simple evidence in favour of specification 2 of the
HNKPC while specifications 1 and 3 seem rejected.

FIGURE 1. ESTIMATIONS OF THE FORWARD AND BACKWARD SHARES OF

THE HNKPC FOR EURO AREA

1
0,9
0,8
Backward share

0,7
0,6 HNKPC estimates
0,5
Mean estimates
0,4
0,3 Median estimates
0,2
0,1
0
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
Forward share

Notes: N=88 estimations from 21 articles; period 2001-2010. Each scatter plot in lozenge form
represents a collected estimation. The x axis shows the forward-looking share of the HNKPC (βf),
while the y axis indicates the backward-looking share (βb). The square form scatter plot is the mean
of the estimates, while that with triangular form is the median.

Data are further analysed with descriptive statistics (Table 2, Figures 2 to 5)26.
The βf median is 0.69, while its mean is 0.68 (Table 2). The distribution is not
symmetric, with a distribution bias on the left (negative Pearson coefficient).
The βb median is 0.252 and the mean 0.278. The distribution bias is on the
contrary on the right (positive Pearson coefficient). Wide variation and
dispersion in estimates is also observed (standard deviation, Figures 2 and 3).
These descriptive statistics are more in line with HNKPC specification 2 than
with specification 3 with a large βf but not equal to one, and βb does not seem
superior to 0.5.

26
As in De Grauwe and Costa-Storti (2004 :5).

344
 EMMANUEL CARRÉ

TABLE 2. EURO AREA HNKPC ESTIMATES: DESCRIPTIVE

STATISTICS
All estimates Estimates respecting the natural rate
βf βb βf βb
Minimum 0,000 0,000 0,401 0,000
Maximum 0,990 0,594 0,990 0,594
Median 0,690 0,252 0,659 0,340
Mean 0,680 0,278 0,667 0,329
Variance 0,019 0,020 0,021 0,021
Standard deviation 0,138 0,142 0,145 0,146
Skewness (Pearson) -1,197 0,129 0,188 -0,226

FIGURES 2 TO 5. HISTOGRAMS OF EURO AREA HNKPC ESTIMATES: ALL

ESTIMATES AND THOSE RESPECTING THE NATURAL RATE

0,5 0,5 0,5 0,5

Frequency
Frequency

0,4 0,4 0,4

Frequency
0,4
Frequency

0,3 0,3 0,3 0,3

0,2 0,2 0,2 0,2
0,1 0,1 0,1 0,1
0 0 0 0
0 0,2 0,4 0,6 0,8 1 0 0,2 0,4 0,6 0,8 1 0 0,2 0,4 0,6 0,8 1 0 0,2 0,4 0,6 0,8 1
Forward share Backward share Forward share Backward share

Notes: Figures 2 and 3 relate to all estimates (88 observations) while Figures 4 and 5 correspond to
estimates respecting the natural rate (39 observations).

As a first simple robustness check of these results, the database sub-sample

composed of estimates respecting the natural rate hypothesis is inspected.

Estimates respecting the natural rate hypothesis. The selection of empirical

estimates respecting the natural rate hypothesis leads to keep 39 estimates from
15 articles with estimation points on the same line (diagonal, Figure 1), because
of the restriction they respect. In the detail of descriptive statistics (Table 2,
Figures 4-5), the βf median is 0.659 with a distribution bias on the right, while
the βb median is 0.34 with a distribution bias on the left. Variation in reported
estimates is also present (Figures 4-5). This corroborates previous indications in
favour of specification 2 with about two-thirds on forward-looking inflation and
one-third on backward-looking inflation. This conclusion is examined via
simple t-tests.

2.2. SIMPLE TESTS ON THE COMPETING SPECIFICATIONS OF THE HNKPC

The competing (H)NKPC specifications are examined using simple t-tests on

collected estimates (Table 3)27. The purely forward-looking NKPC (on average
βf=1) corresponding to specification 1 is not validated. Yet, another simple test
(Table 3, column 2) suggests that βf is not absent. Specification 1 is also
inspected by testing if on average βb=0. Results are not significant, confirming
that the purely forward-looking NKPC is not necessarily an accurate
specification of euro area inflation dynamics. A further t-test examines if, as

27
As in De Grauwe and Costa Storti (2004:6).

345
 THE NEW KEYNESIAN PHILLIPS CURVE FOR THE EURO AREA:
WHICH SPECIFICATION?

defended by specification 2 and suggested by previous evidence, we have mean

βf>βb; this hypothesis is reasserted. Finally, specification 2 is studied more
deeply by testing the HNKPC ‘calibration’ emerging from previous analysis:
βf≈2/3 and βb≈1/3. It is accepted only for estimates respecting the natural rate.
However, the t-test validity can be limited since estimates come from different
econometric methods so that the population homogeneity is not perfect.

TABLE 3. SIMPLE T-TESTS ON HNKPC ESTIMATES FOR THE EURO AREA

Mean Mean Mean Mean βf > Mean Mean
βf =1 βf =0 βb=0 Mean βb βf=2/3 βb=1/3
(t –test) (t -test) (t –test) (t -test) (t –test) (t -test)
All estimates Rejected Rejected Rejected Accepted Accepted Rejected
Estimates respecting the natural rate hypothesis Rejected Rejected Rejected Accepted Accepted Accepted

Note: Acceptation or rejection at the 1% level.

The conclusion of this quantitative review is that specification 2 appears to be

dominant: the HNKPC is preferred to a NKPC, but is mainly forward-looking.
The specification for fractions βf and βb in the euro area seems approximately
two-third/one third:

- βf≈2/3 and βb≈1/3

Although indicative, these results must be received with caution, and cannot
claim to provide a definitive answer to the question of the HNKPC form.
Additional robustness check on related literature and micro-data from firms’
pricing surveys is conducted.

2.3. ROBUSTNESS CHECKS

Related literature. These conclusions are in line with some previous reviews of
the literature on the euro area. McAdam and Willman (2003:37) conclude that
βb is between 0.3 and 0.5, while Altissimo et al. (2006:17) indicate it is ‘less
than 0.5’. Finally, Henzel and Wollmershäuser (2008:823) find that the mean βf
of the studies they review is 0.62, suggesting that βb could be about one third.

Robustness check on micro evidence. (H)NKPC promoters claim it is superior

to previous Phillips curves since it benefits from micro-foundations on firms’
optimising behaviour. Firms’ prices formation on past or future inflation and
information are thus crucial determinants of the HNKPC specification (βf and
βb). As an illustration, the backward-lookingness is introduced via indexation of
firms’ prices on past inflation. The analysis of micro surveys of firms’ pricing
behaviour is therefore a relevant robustness check for the quantitative surveys
results on HNKPC estimates at the aggregate macro level. Firms’ micro
evidence in the euro area confirms previous results of βf≈2/3 and βb≈1/3 since
one of the stylised facts highlighted by micro-level studies is that about 30% or
one third of firms’ prices are indexed on past inflation: ‘it is found that around
one third of firms employ some simple rule of thumb when setting prices’

346
 EMMANUEL CARRÉ

(Alvarez, 2008:23)28. The specification βf≈2/3 and βb≈1/3 seems realistic at the
micro economic level, leading to a macro-micro evidence consistency. A
regression analysis is now undertaken for a deeper investigation of these results.

2.4. REGRESSION ANALYSIS

While the quantitative review clearly signals a substantial backward share, it
also points out an important limitation to this result: significant variations in
reported empirical estimates. A regression on both series of βf and βb estimates
is performed to explore the factors driving these variations (Table 7)29. Debates
in the HNKPC literature direct to three main factors: the estimation technique,
the sample and the driving variable. This statistical analysis offers some rough
indications on their effect on estimates, if they matter for estimates
heterogeneity.

Estimation techniques are evaluated because, as stressed in the ‘intractable

controversy’ discussion, they can be biased, favouring either βf or βb. Four main
estimation methods are distinguished: GMM, ML, Bayesian and TSLS30 (Table
1). A dummy captures the estimation technique by putting 1 when a technique is
used, 0 when it is not. The second factor explaining differences in estimates
could be the sample. Variation in the inflation level can be important for this
variable persistence. Inflation was high in the 1970s in Europe with a shift in
trend inflation, the so called ‘Great Inflation’. That could be a source of inflation
persistence (high βb, low βf)31. On the contrary, a downward shift in trend
inflation is observed in the 1980s32. All samples include the 1980s (Table 1), but
not all include the 1970s. The latter’s effect on estimates is tested. Since the
euro area is analysed, we also test the presence of the post-1999 period in the
data, since the European Monetary Union (EMU) creation could have reduced
inflation persistence (high βf, low βb)33. Yet, this regime shift being recent, it is
only partly included in the samples considered. Hence uncertainty on the
relevance of tests on this period is simply unavoidable. The dummy for the
sample is 1 when the 1970s or the post-1999 period are present in the sample, 0
when they are not. This is an implicit test of the debate on the HNKPC
parameter stability and robustness to the Lucas critique. Thirdly, the literature
informs that the choice of the driving variable serving as indicator of the real
economic activity (x) can impact on estimates: employing an output gap proxy
(detrended output) or instead a marginal cost proxy could make a difference34.
Yet, the expected effect is not clear since while it is generally admitted that the
choice of the output gap impacts on the slope k of the HNKPC, its effect on βf or
βb is more ambiguous. It can lead to higher estimates of βb (lower βf) or the

28
See also Fabiani et al. (2005:11), Alvarez et al. (2005:17), Altissimo et al. (2006: 27) for
confirmation.
29
See De Grauwe and Costa-Storti (2004). I thank the referees for suggesting to investigate further
this analysis.
30
All the Bayesian techniques are grouped together. The same grouping is made for ML and FIML
techniques. The TVC technique is not included in the regression because it concerns only one
estimation.
31
Trichet (2007 :8).
32
See Gali et al. (2001 :1248).
33
Trichet (2007 :8) and Benati (2008 :1018).
34
Gali et al. (2001 :1243).

347
 THE NEW KEYNESIAN PHILLIPS CURVE FOR THE EURO AREA:
WHICH SPECIFICATION?

contrary35. Consequently, it possibly has no significant effect on βf and βb. The

dummy value is 1 when the output gap is used as the driving variable, 0
otherwise. The tested regression on the estimates Z (βf and βb) is in fine36:

Z =intercept + ζTdummy technique + ζS dummy sample + ζOG dummy output gap (3)

This regression could suffer from several limitations. First, it can omit important
factors influencing the estimates, for example the choice of the instrument set
(GMM) or exogenous crucial open economy shock (cost push) or global
structural factors such as the Great Moderation. Secondly, interaction terms are
not considered, so that interaction for instance of the estimation technique and
the driving variable are not studied. Finally, since each estimate is weighted
equally, results of articles with a large number of studies can be outweighed
compared to those with a small number of estimates. For all these reasons the
regression results remain simply an indication and have to be taken carefully, as
underlined by the relatively limited value of the R-squared (Table 7).

TABLE 7. RESULTS OF REGRESSIONS ON THE HNKPC ESTIMATES DATABASE

Regressions: All dummies Regressions: Only significant dummies
Dependent variables (Z)
βf βb βf βb
(A) (B) (C) (D)
Explanatory variables Coefficient S.E. Coefficient S.E. Coefficient S.E. Coefficient S.E.
Intercept 0.919*** 0.152 -0.078 0.14 0.819*** 0.041 -0.134 0.129
Estimation techniques
GMM -0.292* 0,133 0.377** 0.129 -0.126** 0.039 0.373** 0.128
ML -0.354* 0.140 0.444** 0.136 -0.192*** 0.055 0.435** 0.133
TSLS -0.340* 0.153 0.452** 0.148 -0.174* 0.085 0.451** 0.147
Bayesian -0.167 0.134 0.263* 0.130 - - 0.273* 0.129
Driving variable
Output gap 0.003 0.053 -0.042 0.051 - - - -
Sample
Includes the 1970s 0.065 0.064 -0.057 0.062 - - - -
Includes post 1999 -0.070* 0.034 0.136*** 0.033 -0.074* 0.032 0.134*** 0.031
Multiple R² 0.192 0.287 0.167 0.271
Adjusted R² 0.126 0.225 0.126 0.227

This table reports the Ordinary Least Squares linear regression of equation (3).
In columns are indicated the estimates on which the regression is leaded
(dependent variables Z): estimates of the forward-looking share βf of the
HNKPC or estimates of its backward-looking fraction βb. For each of these
estimates, the whole database is used, i.e. N=88 observations for each fraction
for the period 2001-2010. Columns A and B are regressions with all the
dummies. In columns C and D statistically non significant explanatory variables
(dummies) are removed. S.E.: Standard Error. ***, ** and * indicate the
coefficient is statistically significant respectively at the 0%, 0.1% and 1% levels.

We focus on results of regressions with all the dummies (Table 7, columns A

and B) since regressions with only the statistically significant dummies (Table
7, columns C and D) confirm their results. For all explanatory variables the
signs of coefficients are consistent: opposite signs for βf and βb. All statistically

35
Henzel and Wollmershäuser (2008:821), Fanelli (2005:19).
36
It seems valuable to test both βf and βb because not all studies follow the condition βf +βb = 1.

348
 EMMANUEL CARRÉ

significant coefficients point to a negative effect on βf and a positive effect on

βb. Moreover, when comparing the three explanatory variables blocks, the
magnitude of estimated coefficients shows that estimation techniques have more
effect on estimates than the ‘output gap’ or the ‘sample’. All techniques
coefficients have a negative sign for the fraction βf, and on the contrary a
positive sign for βb. It suggests that they could be biased: underestimating βf
while overestimating βb. This bias was expected for the ML technique, but the
opposite was expected for the GMM. The Bayesian technique seems to have
less effect on estimates than other methods in terms of the coefficients values,
but also because one coefficient is not statistically significant (Table 7, column
A). From regression coefficients emerges the following ranking of estimation
technique from the more to the least important effect on estimates: TSLS, ML,
GMM, Bayesian. This reinforces the recent shift in the literature from the
standard GMM to the Bayesian estimation of the HNKPC. The output gap
dummy coefficient is not statistically significant, in accordance with the view
that it impacts more on k than on βf and βb37. The sample dummy is not
statistically significant for the 1970s but is for the ‘post-1999’. In contrast with
common wisdom, there is no clear evidence that the 1970s Great Inflation has
an effect on estimates. On the contrary, the EMU launch seems to matter38, but
not in the direction expected by the ECB: it tends to increase βb instead of βf.
Nevertheless, compared to other factors, this effect is limited, the coefficient on
EMU for βf being more than five time less than the coefficient for ML, and it is
less than 0.1. This small gap with the ECB expectations, and its policy
implications for the conduct of monetary policy, are now to be studied.

3. POLICY IMPLICATIONS

We discuss the implications for the conduct of monetary policy of this result of
a significant βb, i.e. some indications of a HNKPC instead of a NKPC for the
euro area. First, monetary policy under the HNKPC is briefly presented. Next,
the ECB position in favour of the NKPC is examined. Finally, the reasons that
could account for this gap between the quantitative survey results and the ECB
are inspected.

Standard policy under inflation persistence. The quantitative review indicates

that βb>0. The literature documents that this inflation persistence matters for the
conduct of monetary policy because it makes disinflation more costly39 and ‘in
the presence of a large inertial component in inflation, it is instead optimal to
bring inflation down gradually’ (Fuhrer and Olivei, 2004:16)40. More
specifically, in case of uncertainty on the degree of inflation persistence, this
literature recommends to act as if it was important, and to behave as if the
economy was characterized by a HNKPC (Angeloni et al., 2003:545).
ECB members are not perfectly in accordance with the quantitative review
results, since they claim that inflation persistence has now almost disappeared,

37
Jondeau and Le Bihan (2005:536).
38
A very basic indication that the HNKPC parameters are not necessarily perfectly time-invariant.
39
Woodford (2003:499), Jondeau and Le Bihan (2005 :522), Lacker and Weinberg (2007 :221).
40
See also Angeloni et al. (2006:562) on how inflation persistence modifies the ECB interest rate
setting.

349
 THE NEW KEYNESIAN PHILLIPS CURVE FOR THE EURO AREA:
WHICH SPECIFICATION?

so that βb=041. They allege it could have been important in the past, but that it is
not structural since it varies with monetary policy shifts. They say that the EMU
launch has been accompanied by a reduction to zero of inflation persistence:
βb=0. Hence, the euro area inflation dynamics follows a purely forward NKPC,
not a HNKPC. This ECB’s view is consistent with the Gali et al. (2001)
conclusion. In this interpretation, βb>0 signals an imperfectly credible central
bank unable to anchor inflation expectations. On the contrary, a perfectly
credible central bank, intensively using communication, leads firms to form
their prices on the inflation target and in a forward-looking manner, abandoning
indexation on past inflation.

The debate. This ECB case for βb=0 is tempered by the previous quantitative
review indicating instead that βb>0 and that the EMU launch is apparently not
totally encouraging βb=0. This difference could signal that the ECB officials’
approach can be questioned on several grounds. First, the ECB is only a recent
creation, so that it can be premature to conclude that βb=0 is structural.
Secondly, only a few studies42 indicate this result for the EMU era and they are
recent, their conclusions can be thus regarded as tentative, as stressed by ECB
members themselves43. It contrasts with estimates integrated in the quantitative
review that have been intensively debated in the literature. Moreover, some
studies confirm our results that the EMU launch in 1999 had little impact on
inflation persistence: change happened before (Angeloni et al., 2006:382).
Furthermore, the decline in βb can be due to other structural changes in the
economy, to a global exogenous shock such as the Great Moderation or the
reduction of exogenous cost push shock44. The EMU is not necessarily the only
cause, and the causality is more often suggested than formally demonstrated by
ECB members. More generally, as stressed in the introduction, the inflation
persistence decline is not perfectly consensual in the literature. And even if it is
highly plausible that the ECB strategy has reduced inflation persistence, the
claim that it has totally disappeared is far from being consensual.

Thirdly, the ECB explanation of the decline of βb to zero rests upon the elusive
inflation expectations that are difficult to understand. Expectations are crucial
since the contraction of βb defended by the ECB cannot be achieved by the sole
effect of a monetary policy regime shift, for instance with an increase in the
response coefficient on inflation in the monetary policy rule (Benati,
2008:1035). Communication and the expectations management can help to
reduce inflation persistence but it has also to be formally demonstrated that they
can completely eradicate inflation persistence. Some countries under inflation
targeting have a longer experience than the ECB regarding the creation of a new
monetary regime since they began at the beginning of the 1990s. Their
experience stresses that inflation expectations anchorage and reduction cannot
be taken for granted45. Consequently, it appears that a structural change in the

41
Trichet (2007:9), Stark (2008), Tumpel-Gugerell (2008).
42
ECB officials generally only quote Benati (2008). It is mentioned in eight ECB members’
speeches, while the article giving a different view (Angeloni, Aucremanne and Ciccarelli, 2006) is
mentioned once.
43
Stark (2008).
44
This is mentioned in Angeloni et al. (2006:366), Benati (2008:1040).
45
Lin and Ye (2007:2529).

350
 EMMANUEL CARRÉ

economy could be required on top of the policy change to induce the move to
the canonical NKPC.

Fourth, the transition to βb=0 can be expected to be seen at the micro-level of

firms’ price formation. If their expectations evolve from backward to forward-
looking, it should impact their pricing rule. Microeconomic surveys of firms
prima facie do not support this idea since they show that firms’ price behaviour
has not radically change next to the EMU launch (Angeloni et al., 2006:366).
Yet, this is indirect evidence on their pricing rule, but to our knowledge there is
no study of the evolution, before and after EMU, of the firms’ share indexing
their prices on past inflation. Nevertheless, we highlighted that these surveys
report a significant share of backward-looking price setters. Hence, compared to
our results, the micro-macro consistency of the purely forward NKPC for the
euro area is less obvious or, at least, needs more empirical investigation.
Finally, the analysis has stressed that not only the sample, but also the
estimation technique can influence estimates. Thus studies showing that βb=0
have to carefully demonstrate that there is no limitation to the estimation
technique they use, and it is not always the case in the state of art of this
literature that generally does not discuss the Bayesian estimation for instance.
For all these reasons, it is difficult to consider that uncertainty on inflation
persistence is nil, implying that the central bank could act as if it was a fairly
important phenomenon. The ECB should avoid the ‘risk of conducting monetary
policy under the assumption that inflation is not structurally persistent when it
is’ (Benati, 2008:1008). Hence βb>0 is consistent with a conservative and
cautious central banker. The value of βb emerging from the quantitative review
could constitute this conservative calibration of inflation persistence.

CONCLUSION AND DIRECTIONS FOR FUTURE RESEARCH

According to common wisdom, the purely forward-looking NKPC is robustly

based on micro-foundations and capable of reproducing the empirical dynamics
of inflation observed in the euro area. As a first contribution of this article,
stylised facts on empirical estimates of the HNKPC, highlighted by the
quantitative review, indicate that this perceived wisdom has to be taken
carefully. It suggests that the NKPC is empirically more hybrid than pure, as it
contains a backward looking fraction: approximately two-thirds forward and
one-third backward. This result is roughly in agreement with firms
microeconomic surveys, also indicating that about one-third of firms form their
prices based on past inflation. Nevertheless, several limitations qualify this
result and give some directions for future research. Due to strict and numerous
selection criteria, the estimates number is limited. For future research, the
database can be extended to other price setting microfoundations such as Taylor
(1980) that is often judged to lead a NKPC similar to the Calvo pricing. Another
avenue for future research is to examine in more details the various marginal
cost proxies. Estimates from HNKPC specifications incorporating open
economy elements, at least implicitly, could also be included.
Even if these results of the quantitative review should be received with caution,
they propose a substantial departure from the optimal monetary policy derived
from the purely forward looking NKPC. In particular, the Divine Coincidence

351
 THE NEW KEYNESIAN PHILLIPS CURVE FOR THE EURO AREA:
WHICH SPECIFICATION?

and the expectations management – the cornerstones of the ECB priority on

price stability – are not fully valid under a hybrid NKPC. Empirical results
invite to a more flexible monetary policy, more gradual, with a longer optimal
time horizon.

Another cautionary note to this otherwise clear diagnostic in favour of the

HNKPC is the presence of substantial variations in reported estimates. The
second contribution of this article is to explore these variations with a regression
analysis, which results could be of interest for researchers in their future
methodological choice of the HNKPC estimation. It reveals that the estimation
techniques have a significant effect on estimates; the Bayesian method currently
used in DSGE models having the weaker effect. On the contrary, the use of
output gap proxies instead of marginal cost proxies does not have a significant
effect. Finally, in comparison to estimation techniques, the sample employed
has apparently little or no effect on estimates. Samples containing post-1999
data have a little but significant positive effect on βb (negative on βf). The ECB
is expecting the opposite effect from its creation: a reduction of βb. This gap
stressing uncertainty on inflation persistence (approximated by βb) is a subject
for future research on the recent EMU period to obtain more robust analysis of
the effect of its launch on the HNKPC. This also requires more investigations at
the firms’ micro-level to see how the EMU has affected their price-setting
behaviour. In the current state of the art, characterized by uncertainty on
inflation persistence, a large part of the literature recommends the central bank
to overestimate inflation persistence. This leads to a policy environment close to
the HNKPC enlightened by our quantitative review.

352
 EMMANUEL CARRÉ

REFERENCES
Altissimo, F., M. Ehrmann and F. Smets, 2006. “Inflation Persistence and
Price-Setting Behaviour in the Euro Area”, European Central Bank Occasional
Paper, No.46.
Álvarez, L., 2008. “What do Micro Price Data Tell Us on the Validity of the
New Keynesian Phillips Curve”, Economics: the Open-Access, Open
Assessment E-Journal, Vol.2, 2008-19.
Álvarez, L., E. Dhyne, M. Hoeberichts, C. Kwapil, H. Le Bihan, P.
Lünnemann, F. Martins, R. Sabbatini, H. Stahl, P. Vermeulen and J.
Vilmunen, 2005. “Sticky prices in the euro area: a summary of new micro
evidence”, European Central Bank Working Paper, No.563.
Angeloni, I., L. Aucremanne and M., Ciccarelli, 2006. “Price Setting and
Inflation Persistence: Did EMU Matter?”, Economic Policy, 21(46), pp. 354-
387.
Angeloni, I., L. Aucremanne, M. Ehrmann, J. Gali, A. Levin and F., Smets,
2006. “New evidence on inflation persistence and price stickiness in the euro
area: implications for macro modelling”, Journal of the European Economic
Association, 4(2-3), pp. 562-574.
Angeloni, I., G. Coenen and F. Smets, 2003. “Persistence, the Transmission
Mechanism and Robust Monetary Policy”, Scottish Journal of Political
Economy, 50(5), pp.527-549.
Bardsen, G., E. Jansen and R. Nymoen, 2002. “The Empirical (Ir)relevance of
the New Keynesian Phillips Curve”, Norwegian University of Science and
Technology, Working Paper Series, No.21/2002.
Bardsen, G., E. Jansen and R. Nymoen, 2004. “Econometric Evaluation of the
New Keynesian Phillips Curve”, Oxford Bulletin of Economics and Statistics,
66, pp. 671-686.
Barkbu, B., V. Cassino, A. Gosselin-Lotz and L. Piscitelli, 2005. “The New
Keynesian Phillips Curve in the United States and the Euro Area: Aggregation
Bias, Stability and Robustness”, Bank of England Working paper, No.285.
Benati, L., 2008. “Investigating Inflation Persistence Across Monetary
Regimes”, Quarterly Journal of Economics, 123(3), pp.1005-1060.
Calvo, G., 1983. “Staggered Prices in a Utility-Maximizing Framework”,
Journal of Monetary Economics, 12(3), pp. 983–998.
De Grauwe, P. and C. Costa Storti, 2004. “The Effect of Monetary Policy: A
Meta-Analysis”, CESIfo Working Paper, No.1224.
De Walque, G., F. Smets and R. Wouters, 2004. “Price Setting in General
Equilibrium: Alternative Specifications”, presented at the conference Inflation
Persistence in the Euro Area, European Central Bank, 10-11 December.
De Walque, G., F. Smets and R. Wouters, 2006a. “Price Shocks in General
Equilibrium: Alternative specifications”, CESifo Economic Studies, 52(1), pp.
153-176.
De Walque, G., F. Smets and R. Wouters, 2006b. “Firm-Specific production
Factors in a DSGE Model with Taylor Price Setting”, National Bank of
Belgium, Working Paper No.85.
Fabiani, S., M. Druant, I. Hernando, C. Kwapil, B. Landau, C. Loupias, F.
Martins, T. Mathä, R. Sabbatini, H. Stahl, and A. Stokman, 2005. “The
Pricing Behaviour of Firms in The Euro Area: New Survey Evidence”,
European Central Bank Working Paper, No.535.

353
 THE NEW KEYNESIAN PHILLIPS CURVE FOR THE EURO AREA:
WHICH SPECIFICATION?

Fagan, G., J. Henry, and R. Mestre, 2001. “An Area-Wide Model (AWM) for
the Euro Area”, European Central Bank Working Paper, No.42.
Fanelli, L., 2005. “Testing the New Keynesian Phillips Curve through Vector
Autoregressive Models: Results from the Euro Area”, MPRA Paper, No.2380.
Fuhrer, J. and G. Olivei, 2004. “Estimating Forward-Looking Euler Equations
with GMM Estimators: an Optimal Instruments Approach”, Federal Reserve
Bank of Boston Working paper, No.04-2.
Fuhrer, J., 1997. “The (Un)importance of Forward-Looking Behaviour in Price
Specification”, Journal of Money, Credit and Banking, 29(2), pp. 338-350.
Gagnon, E. and H. Khan, 2001. “New Phillips Curve with Alternative
Marginal Cost Measures for Canada, the United States, and the Euro Area”,
Bank of Canada Working Paper, No.2001-25.
Gagnon, E. and H. Khan, 2005. “New Phillips Curve under Alternative
Production Technologies for Canada, the United States, and the Euro Area”,
European Economic Review, 49(6), pp.1571-1602.
Gali, J. and M. Gertler, 2007. “Macroeconomic Modelling for Monetary
Policy Evaluation”, Journal of Economic Perspectives, 21(4), pp. 25-45
Gali, J., M. Gertler and J.D. Lopez-Salido, 2001. “European Inflation
Dynamics”, European Economic Review, 45(7), pp. 1237–1270.
Gali, J., M. Gertler, and J.D. Lopez-Salido, 2003. “Erratum to European
Inflation Dynamics”, European Economic Review, 47(4), pp. 759–760.
Goodfriend, M., 2007. “How the World Achieved Consensus on Monetary
Policy”, Journal of Economic Perspectives, 21(4), pp. 47-68.
Groen, J. and H. Mumtaz, 2008. “Investigating the Structural Stability of the
Phillips Curve Relationship”, Bank of England Working Paper, No.350.
Henzel, S. and T. Wollmershäuser, 2008. “The New Keynesian Phillips Curve
and the Role of Expectations: Evidence from the Ifo World Economic Survey”,
Economic Modelling, 25(5), pp.811-832.
Hondroyiannis, G., P. Swamy and G. Tavlas, 2008. “Inflation Dynamics in
the Euro Area and in New EU Members: Implications for Monetary Policy”,
Economic Modelling, 25(6), pp.1116-1127.
Jondeau, E. and H. Le Bihan, 2005. “Testing for a Forward-Looking Phillips
Curve. Additional Evidence from European and US Data”, Economic
Modelling, 22(3), pp. 521-550.
Jondeau, E. and H. Le Bihan, 2008. “Examining Bias in Estimators of Linear
Rational Expectations Models under Misspecification”, Journal of
Econometrics, 143(2), pp. 375-395.
Jondeau, E. and J.-G. Sahuc, 2008. “Optimal Monetary Policy in an Estimated
DSGE Model of the Euro Area with Cross-Country Heterogeneity”,
International Journal of Central Banking, 4(2), pp. 23-72.
Kleibergen, F. and S. Mavroeidis, 2009. “Weak Instrument Robust Tests in
GMM and the New Keynesian Phillips Curve”, Journal of Business &
Economic Statistics, 27(3), pp.293-311.
Lacker, M. and J. Weinberg, 2007. “Inflation and Unemployment: a
Layperson’s Guide to the Phillips Curve”, Federal Reserve Bank of Richmond
Economic Quarterly, 93(3), pp.201-227.
Lin, S. and H. Ye, 2007. “Does Inflation Targeting Really Make a Difference?
Evaluating the Treatment Effect of Inflation Targeting in Seven Industrial
Countries”, Journal of Monetary Economics, 54(8), pp. 2521-2533.

354
 EMMANUEL CARRÉ

Lindé, J., 2005. “Estimating New Keynesian Phillips Curves: a Full

Information Maximum Likelihood Approach”, Journal of Monetary Economics,
52(6), pp. 1135-1149.
Lippi, F. and S. Neri, 2007. “Information Variables for Monetary Policy in an
Estimated Structural Model of the Euro Area”, Journal of Monetary Economics,
54(4), pp.1256-1270.
Ma, A., 2002. “GMM Estimation of the New Keynesian Phillips Curve”,
Economics Letters, 76(3), pp. 411–417.
Mankiw, G., 2001. “The Inexorable and Mysterious Trade-Off between
Inflation and Unemployment”, The Economic Journal, 111(471), pp. 45-61.
Mavroeidis, S., 2005. “Identification Issues in Forward-Looking Models
Estimated by GMM, with an Application to the Phillips Curve”, Journal of
Money, Credit and Banking, 37(3), pp. 421-448.
McAdam, P. and A. Willman, 2003. “New Keynesian Phillips Curves: a
Reassessment Using Euro-Area Data”, European Central Bank Working Paper,
No.265.
McCallum, B., 1997. “Comments on an Optimization-Based Econometric
Framework for the Evaluation of Monetary Policy by Rotemberg, J. and M.,
Woodford”, in B. Bernanke and J. Rotemberg, (eds), NBER Macroeconomics
Annual, Cambridge, MA: MIT Press, pp. 355-359.
McCallum, B., 2008. “Reconsideration of the P-bar Model of Gradual Price
Adjustment”, European Economic Review, 52(8), pp1480-1493.
Onatski, A. and N. Williams, 2010. “Empirical and Policy Performance of a
Forward-Looking Monetary Model”, Journal of Applied Econometrics, 25(1),
pp.145-176.
O’Reilly, G. and K. Whelan, 2004. “Has Euro-Area Persistence Changed Over
Time?”, European Central Bank Working paper, No. 335.
Paustian, M. and E. Pytlarczyk, 2006. “Sticky Contracts or Sticky
Information? Evidence from an Estimated Euro Area DSGE”, Presented at the
Bank of Canada Conference ‘Dynamic Models Useful for Policy Making’,
September 8-9.
Phillips, A., 1958. “The Relationship between Unemployment and the Rate of
Change of Money Wages in the United-Kingdom 1861-1957”, Economica,
25(100), pp. 283-299.
Roberts, J., 1995. “New Keynesian Economics and the Phillips Curve”, Journal
of Money, Credit and Banking, 27(4), pp. 975-84.
Rudd, J. and K. Whelan, 2005. “New Tests of the New Keynesian Phillips
Curve”, Journal of Monetary Economics, 52(6), pp. 1167-1181.
Rudd, J. and K. Whelan, 2007. “Modeling Inflation Dynamics: a Critical
Review or Recent Research”, Journal of Money, Credit and Banking, 39(1), pp.
155-170.
Rumler, F., 2005. “Estimates of the Open Economy New Keynesian Phillips
Curve for Euro Area Countries”, European Central Bank Working paper,
No.496.
Sahuc, J.G. and F. Smets, 2008. “Differences in Interest Rate Policy at the
ECB and the Fed: An Investigation with a Medium-Scale DSGE Model”,
Journal of Money, Credit and Banking, 40(2-3), pp.505-521.
Smets, F. and R. Wouters, 2003. “An Estimated Dynamic Stochastic General
Equilibrium Model of the Euro Area”, Journal of the European Economic
Association, 1(5), pp. 1123-1175.

355
 THE NEW KEYNESIAN PHILLIPS CURVE FOR THE EURO AREA:
WHICH SPECIFICATION?

Smets, F. and R. Wouters, 2004. “Forecasting with a Bayesian DSGE model:

An application to the Euro Area”, Journal of Common Market Studies, 42(4),
pp. 841-867.
Stark, J., 2008. “Lessons for Central Bankers from the history of the Phillips
Curve”, speech, ECB, June 11.
Taylor, J., 1980. “Aggregate Dynamics and Staggered Contracts”, Journal of
Political Economy, 88(1), pp. 1-23.
Tempel-Gugerell, G., 2008. “Challenges for the Euro at Ten”, Speech, ECB,
December 8.
Trichet, J.C., 2007. “The Euro Area and Its Monetary Policy”, Address at the
conference ‘The ECB and Its Watchers’, Frankfurt am Ain, 7 September.
Woodford, M., 2003. Interest and Prices: Foundations of a Theory of
Monetary Policy, Princeton, NJ: Princeton University Press.

356