Gradualism vs Cold Turkey

The paper analyzes the incentive for the ECB to establish
reputation by pursuing a restrictive policy right at the start of
its operation. The bank is modelled as risk averse with respect to
deviations of both inflation and output from her target. The
public, being imperfectly informed about the bank’s preferences
uses observed inflation as (imperfect) signal for the unknown
preferences. Under linear learning rules - which are commonly used
in the literature - a gradual build up of reputation is the optimal
response. The paper shows that such a linear learning rule is not
consistent with efficient signaling. It is shown that in a game
with efficient signaling, a cold turkey approach - allowing for
deflation - is optimal for a strong bank - accepting high current
output losses at the beginning in order to demonstrate its
toughness.