Identifying Monetary Policy Shocks: A Natural Language Approach

This paper proposes a novel method for the identification of monetary policy shocks. By applying natural language processing techniques to documents that staff economists at the Federal Reserve prepare for FOMC meetings, we capture the information set of the committee at the time of policy decisions. We verify econometrically that the language contains valuable information beyond what is incorporated in the staff's numerical forecasts. Using machine learning techniques, we then predict changes in the target interest rate conditional on the committee's information set and obtain a measure of monetary policy shocks as the residual. We find that the dynamic responses of macro variables to our identified shocks are consistent with the theoretical consensus.