Unveiling the Federal Open Market Committee's Intent Through NLP Analysis of FOMC Speeches: leveraging text analytics to decode future Federal Funds Rate target range decisions

Abstract

This research paper examines whether the sentiments and topics within speeches made by the Federal Open Market Committee during inter-meeting periods can tip off investors about future rate decisions made by the FOMC in changing the Federal Funds Rate target range. 1066 speeches made by Federal Open Market Committee members from 2006 to 2024 are assessed, obtained from the Federal Reserve’s website using a data scraper developed in Python. Subsequently, the speech dataset is transformed into a corpus for Natural Language Processing. Sentiment score variables are computed utilizing the Loughran-MacDonald financial sentiment dictionary, as well as topic mention counts, computed using a similarity function which leverages a similarity function based upon FinText’s financial word-embeddings designed by Rahimikia (2021). After controlling for Taylor Rule (1993) forward-looking macroeconomic indicators, ordered probit regression models are implemented following Hayo & Neuenkirch (2010) to assess the sentiment and topic variables’ efficacy in predicting future FOMC target rate decisions. It is found that increasing levels of negative sentiment and increasing mentions of lending and borrowing topic-related terms within inter-meeting FOMC speeches indicate a decreased probability of FOMC target rate hikes at the next upcoming FOMC meeting. These results align with the findings of Hayo & Neuenkirch (2010), further suggesting that leveraging FOMC inter-meeting communications provide valuable insights into future FOMC target rate decisions beyond what is implied by Taylor Rule parameters.