How Cenral Banks Learn the True Model of the Economy

Abstract

Policy decisions affect economic outcomes, and the likelihood of observing a given state of the world. We investigate how policy choices affect learning of the true model of the economy when the policymaker’s model is mis-specified. We ask under what conditions can the central bank learn the correct specification of the model describing the economy, and what is the impact of exogenous shocks and of adopting an optimal monetary policy on the speed of learning. Slow learning can occur simply because identifying the correct model at standard confidence levels requires a long data sample. We show that neither real-time learning by the policymaker or the private sector, nor the adoption of an optimal policy, affect the speed of detection of model misspecification. Detection speed depends instead on the relative volatility of supply and demand shocks.