New Keynesian Explanations of Cyclical Movements in Aggregate Inflation and Regional Inflation Differentials
ABSTRACT What determines the cyclical behavior of aggregate inflation and regional inflation differentials? The answer has strong implications for monetary policy and in Europe for the Stability and Growth Pact. In the United States, inflation rates move pro-cyclically, and across the Euro Area, inflation differentials are positively correlated with growth differentials. This suggests that demand shocks are the primary determinants of the cyclical behavior of aggregate inflation and regional inflation differentials. In this paper, we discuss New Keynesian explanations of these correlations, and we argue that demand shocks are either missing or inadequately modeled in the in typical New Keynesian model. Copyright Springer Science + Business Media, Inc. 2006
New Keynesian Explanations of
Cyclical Movements in Aggregate Inflation
and Regional Inflation Differentials
Matthew B. Canzoneri, Robert E. Cumby, Behzad T. Diba and Olena Mykhaylova
Economics Department, Georgetown University
First Draft: June 7, 2005
What determines the cyclical behavior of aggregate inflation and regional inflation
differentials? The answer has strong implications for monetary policy and in Europe
for the Stability and Growth Pact. In the United States, inflation rates move pro-
cyclically, and across the Euro Area, inflation differentials are positively correlated
with growth differentials. This suggests that demand shocks are the primary
determinants of the cyclical behavior of aggregate inflation and regional inflation
differentials. In this paper, we discuss New Keynesian explanations of these
correlations, and we argue that demand shocks are either missing or inadequately
modeled in the in typical New Keynesian model.
Key Words: inflation, inflation differentials, NNS models
JEL Classification: E10, E31, E63
1 Fatas and Mihov (2001a ,b), Blanchard and Perotti (2002) and Canzoneri, Cumby and
Diba (2002) find that an increase in government spending increases consumption and output;
Canzoneri, Cumby and Diba (2002) also find that the Federal Funds rate reacts in a manner that
is consistent with standard Henderson-McKibbin-Taylor rules. On the other hand, Perotti (2004)
suggests that the effect on consumption may have diminished in recent years, and he questions
whether it exists at all in some European countries.
2 For HP-filtered quarterly U.S. data from 1960.1 to 2003.2, the correlation between CPI
inflation and the log of GDP is 0.33; the correlation between the Federal Funds Rate and the log
of GDP is 0.35. These correlations do not change sign for leads or lags of one quarter. See the
appendix of CC&D for data sources.
Traditional Keynesian explanations of the cyclical movements of inflation focused primarily
on demand shocks. An increase in say government spending was thought to have multiplier effects
on consumption and output, and the increase in aggregate demand would eventually create inflation
(via the Phillips Curve) and an increase in the interest rate (via the central bank’s monetary policy
rule). U.S. data appear to be consistent with this view: recent VAR studies suggest that con-
sumption rises in response to a government spending shock, and that the Federal Funds Rate rises
in response to the increase in output and inflation;1 moreover, the unconditional correlation between
inflation and output is positive (0.33), and so is the correlation between nominal interest rates and
output (0.35).2 Recent data from the Euro Area also appear to be consistent with the traditional
Keynesian view: national inflation differentials are positively correlated with national growth
differentials (a fact we will document below).
The Real Business Cycle (RBC) model that followed focused primarily on productivity
shocks. In the RBC view, productivity shocks drive fluctuations in output, while the cyclical
behavior of interest rates and inflation is simply the manifestation of a monetary policy that is
otherwise irrelevant. More recently, a New Neoclassical Synthesis (NNS) adds monopolistic
competition and nominal inertia to the RBC model to create a new Keynesian model in which both
3 Goodfriend and King (1997) outlined the New Neoclassical Synthesis, and gave it the
name. Woodford (2003) provides a masterful introduction to this class of models. NNS models
are now being used widely in the academia and at policy making institutions. Important early
contributions to the study of monetary policy include Rotemberg and Woodford (1997), King
and Wolman (1999), and Erceg, Henderson and Levin (2000). Recent extensions to include
fiscal policy include Benigno and Woodford (2003) and Schmitt-Grohe and Uribe (2004).
Larger institutional models include the Bank of England’s BEQM (see Bank of England (2004)),
the IMF’s GEM (see Bayoumi et al (2004)), and the FRB’s SIGMA (see Erceg et al (2004));
similar models are being developed at the ECB and a number of other central banks.
productivity shocks and demand shocks play a role in the cyclical movements of interest rates and
inflation.3 In NNS models, demand shocks tend to produce procyclical movements in interest rates
and inflation, while productivity shocks tend to produce countercyclical movements.
In this paper, we analyze standard NNS models to see if they are capable of generating the
procyclical movements of interest rates and inflation that are observed in the data. We begin with
a model developed in Canzoneri, Cumby and Diba (CC&D) (2004). The CC&D model captures
some key features of the U.S. business cycle, but as we shall see it generates strongly negative
correlations between interest rates and output, and between inflation and output. We attribute this
model failure to the fact that – despite the presence of shocks to government purchases and the
interest rate rule – productivity shocks play a dominant role in the determination of inflation:
variance decompositions indicate that productivity shocks explain 95% of the fluctuations in
inflation in the CC&D model. We suspect that some demand side shocks are either absent or
incorrectly modeled, and we investigate both possibilities in this paper.
We begin by augmenting the CC&D model with a private spending shock that has been
suggested by Ireland (2004), Gali and Rabanal (2004) and others. Private spending shocks – like
government spending shocks – produce procyclical movements in interest rates and inflation, and
this increases the unconditional correlations of interest rates and inflation with output. However,
4 Some of the larger institutional models employ a similar device.
private spending shocks are modeled as shocks to preferences, and – unlike government spending
shocks – they are not directly observable: it is unclear how large we can plausibly make them. In
the CC&D model, for standard deviations consistent with the existing literature, the unconditional
correlations of interest rates and inflation with output remain negative.
For this reason, we go on to investigate the possibility that the propagation of fiscal shocks
is incorrectly modeled in the typical NNS model. The CC&D model is Ricardian in the sense that
households respond to an increase in government spending (and the implied increased tax burden)
by working more and spending less, in apparent contradiction to the VAR studies cited earlier. This
raises the possibility that a government spending shock has less effect on aggregate demand in the
model than it does in the U.S. economy. Galì, López-Salido and Vallés (2004) have shown that
adding “rule of thumbers” – households that just consume their income each period – can make
aggregate consumption rise in response to an increase in government spending.4 Here, we add rule
of thumbers to the CC&D model to see if we can generate the procyclical movements in inflation
and interest rates that are observed in the data.
The CC&D model describes a single country with a single aggregate production sector. In
this paper, we extend the CC&D model to a two country currency union, and we investigate its
explanation of the cyclical behavior of the national inflation differentials.
The early experience of the Euro has generated interest in explanations of national inflation
differentials. Differences between national inflation rates and the Euro area average are proving to
5 ECB (2003) documents these inflation differentials. Altissimo, Benigno and Rodriguez
Palenzuela (2004) provide an interesting statistical decomposition of the inflation differentials.
See also Duarte (2003) and Angeloni and Ehrmann (2004).
6 Quarterly inflation differentials for country J are computed as 4(log(PJ,t/PJ,t-1) -
log(PE,t/PE,t-1)), where PJ,t is the average over the three months of quarter t of the HIPC for
country J and PE,t is similarly defined for the Euro Area. Real growth differentials are computed
similarly by taking annualized averages of quarterly growth rates of real GDP and subtracting
the annualized average quarterly growth rate for the Euro Area. The source for both the HIPC
and real GDP data is Eurostat.
be larger than many had anticipated.5 Figure 1 illustrates the average inflation differentials since the
Euro’s inception;6 they range from a high of 1.8% p.a. in Ireland to a low of - 0.6% p.a. in Germany.
These inflation differentials are also quite volatile; for example, the standard deviation of the
inflation differential between France and Germany is 1.6% p.a.
We are not aware of any rigorous analysis of the welfare consequences of these inflation
differentials, but the way in which they are being viewed seems to depend upon what is thought to
be generating them. When the differentials are thought to be driven by unstable fiscal policies, then
the presumption seems to be that the Stability and Growth Pact may be useful in controlling them.
When the differentials are thought to be driven by other national or regional demand disturbances,
then the presumption seems to be that the Stability and Pact is getting in the way of automatic
stabilizers embodied in national fiscal policies. And finally, when the differentials are thought to
be driven by asymmetric productivity shocks, the presumption seems to be that the differentials
reflect relative price movements that do not need to be corrected. While a rigorous welfare analysis
is well beyond the scope of this paper, it is clearly of interest to ask what is driving the inflation
differentials, both in the data and in the NNS models that are currently being used to evaluate policy.
Figure 1 illustrates what might be described as a cross-sectional Phillips Curve for the Euro