Central Bank Language Modeling to Predict Changes in Market Interest Rate Expectations

Abstract

This study seeks to predict the impact of central bank communication on market interest rate expectations with the goal of enabling more finely tuned and informed central bank communication that can avoid misunderstandings with market participants. We consider post-meeting statements released by the Federal Reserve's Federal Open Market Committee. These statements are manually assigned a topic and polarity label, and an automated labeling scheme grounded in classification methods and dimensionality reduction techniques is constructed. Changes in market interest rate expectations are understood through price changes of U.S. Treasury futures contracts. To predict price changes, several regression techniques are used with the FOMC statement data and auxiliary economic variables as input. The price changes are also converted to categorical variables for class prediction. This study serves as an attempt at improving central bank communication and as an addition to scholarly research concerned with the economic and financial impact of central bank communication.