On the perils of stabilizing prices when agents are learning

We show that price level stabilization is not optimal in an economy where agents have incomplete knowledge about the policy implemented and try to learn it. A systematically more accommodative policy than what agents expect generates short term gains without triggering an abrupt loss of confidence, since agents update expectations sluggishly. In the long run agents learn the policy implemented, and the economy converges to a rational expectations equilibrium in which policy does not stabilize prices, economic volatility is high, and agents suffer the corresponding welfare losses. However, these losses are outweighed by short term gains from the learning phase.