894 Macroeconometric Modeling for Policy
e.g., pre-whitening switched on in generalized method of moments (GMM), and
leading papers downplay the relevance of congruency for the evaluation of the
NKPC.
After assessing some of the critiques that have been directed towards the NKPC,
Gali, Gertler and López-Salido (2005) assert that the NKPC, in particular the domi-
nance of forward-looking behavior, is robust to the choice of estimation procedure
and to possible specification bias. They conclude that the following three results
are proven characteristics of NKPC for all datasets:
- The two null hypothesesaf=0 andab=0 are rejected both individually and
jointly. - The coefficient on expected inflation exceeds the coefficient on lagged inflation
substantially. The hypothesis ofaf+ab=1 is typically not rejected at conven-
tional levels of significance, although the estimated sum is usually a little less
than one numerically.- When real marginal costs are proxied by the log of the wage share, the coeffi-
cientbis positive and significantly different from zero at conventional levels of
significance.
As mentioned, critics of the NKPC have challenged the robustness of all three, but
with different emphases and from different perspectives. The inference procedures
and estimation techniques used by GG and GGL (2001) have been criticized by
Rudd and Whelan (2005, 2007) and others, but GGL (2005) show that their initial
Results 1 and 2 remain robust to these objections.
When it comes to Result 3, GGL (2005) overlook that several researchers have
been unable to confirm their view that the wage share is a robust explanatory
variable in the NKPC. Bårdsen, Jansen and Nymoen (2004) showed that the sig-
nificance of the wage share in the GGL (2001) model is fragile, as it depends on
the exact implementation of the GMM estimation method used, thus refuting that
Result 3 is a robust feature of NKPC estimated on euro-area data.
Fanelli (2008), using a vector autoregressive regression model on the euro-area
dataset, finds that the NKPC is a poor explanatory model. On US data, Mavroeidis
(2006) has shown that real marginal costs appears to be an irrelevant determinant
of inflation, confirming the view in Fuhrer (2006) about the difficulty of obtaining
a sizeable coefficient on the forcing variable in the US NKPC. Already the studies
cited represent evidence that refutes the claim that Result 3 is robust. Instead it is
to be expected that, depending on the operational definition of real marginal costs,
the estimation method and the sample, the numerical and statistical significance
ofbwill vary across different studies.
Of course, Result 3 is just as important as Results 1 and 2 for the status of the
NKPC as an adequate model, so if that part of the model is non-structural, it might
be that that Results 1 and 2 have another explanation than the intended, which is
that there is a good match between the NKPC and the true inflation process. Bård-
sen, Jansen and Nymoen (2004) (euro-area) and Bjørnstad and Nymoen (2008)