Monetary policy analysis in real-time: Vintage combination from a real-time dataset

This paper provides a general strategy for analyzing monetary policy in real time which accounts for data uncertainty without explicitly modelling the revision process. The strategy makes use of all the data available from a real-time data matrix and averages model estimates across all data releases. Using standard forecasting and policy models to analyze monetary authorities' reaction functions, we show that this simple method can improve forecasting performance and provide reliable estimates of the policy model coefficients associated with small central bank losses, in particular during periods of high macroeconomic uncertainty.