Clσ2j , and the delta method to obtain the standard errors. Table 7 displays the results.
The average inflation bias, which is reported in the second row, is sizable and statistically
different from zero only in the pre-Volcker period. The model-based inflation mean in the
fourth row confirms that we effectively decompose the actual inflation mean into a target and
a bias argument. Moreover, a shift in the policy preferences on output stabilization appears
to account for a larger fraction of the difference in the sub-samples mean of inflation relative
to a reduction in the implicit inflation target.
The results in Table 6 and Table 7 suggest that while a different interest rate response
to the inflation level, as described by the rise of c1 , and a more favorable macroeconomic
environment, as summarized by the decline in the standard deviation of the output gap, have
also played a role, a change in the policy preference on output from asymmetric to symmetric
appears crucial to account for the observation that US inflation has been on average higher
during the 1960s and 1970s than during the 1980s and 1990s.
5 Conclusions
The contribution of this paper is twofold. At the theoretical level, it derives the analytical so-
lution of the central bank optimization problem when the policy preferences are asymmetric in
both inflation and output gaps, and the monetary transmission mechanism is New-Keynesian.
The specification of the policy objectives is general enough to nest the quadratic form as a
special case and therefore it translates into a potentially nonlinear targeting rule. This fea-
ture forms the basis of our hypothesis testing for the presence of asymmetric preferences as it
allows to reversely engineer potential evidence of nonlinearities in the reaction function into
evidence of asymmetries in the policy objective.
At the empirical level, this paper shows that US monetary policy can be effectively char-
acterized by a nonlinear policy rule only during the pre-Volcker regime, with the interest rate
response to the state of the business cycle being the dominant type of nonlinearity. In par-
ticular, the Fed appears to have historically attached a larger weight to output contractions
than to output expansions of the same magnitude such as to induce an average inflation bias
of 1.11%. The latter can account for a sizable fraction of the inflation rise observed during
the 1960s and 1970s. These findings are robust across alternative measures of inflation and
output gap, as well as across alternative estimation strategies.
17