representation (6).
3.5 Discussion
It is useful at this point to compare our estimates with the results from some recent studies
that also focus on the policy regime shift of 1979. Clarida, Gali' and Gertler (2000) estimate a
forward-looking linear reaction function for the pre-Volcker period and report values of 0.68 for
the coefficient on inflation (s.e.= 0.06) and 0.28 for the coefficient on CBO output (s.e.= 0.08).
Their estimates suggest that neglecting the squared output gap, which significantly enters
our empirical specification with a negative sign, introduces a downward bias in the linear
estimate.8 Turning to the nonlinear specifications, Dolado, Maria-Dolores and Ruge-Murcia
(2004) use a Clarida, Gali' and Gertler-type of rule augmented with a generated regressor for
the conditional variance of inflation and find no evidence for this form of nonlinearity. Kim,
Osborn and Sensier (2004) use a semi-parametric method of estimation and show that only
the asymmetry over the output gap has been sizable.
The post-Volcker estimates of the parameters on the inflation level and the output gap
level are not statistically different from the values reported in Clarida, Gali' and Gertler (2000),
and therefore they confirm a limited role for nonlinearity during the last two decades. These
results are consistent with those in Kim, Osborne and Sensier (2004) while they are only
marginally so with those in Dolado, Maria-Dolores and Ruge-Murcia (2004). The absence of
an output gap ob jective in the latter study however seems a natural candidate to explain the
difference. Lastly, we line up with earlier contributions in that the coefficient on the inflation
level becomes bigger than one moving from the pre- to the post-1979 period.
4Theaverageinflation bias
The estimates of the previous section support the notion of a novel inflation bias due to Cukier-
man (2002). In the presence of an asymmetric ob jective over the output gap and uncertainty
about the state of the economy, the monetary authorities face an incentive to respond more
aggressively to output contractions of a given amount than to output expansions of the same
magnitude. The reason is that the expected marginal benefit of a policy intervention is convex
in the output gap, implying that to satisfy the Euler equation and stimulate future aggregate
8 This result holds true also for the alternative measures of inflation and output gap.
14