How does the stock market listen to Fedspeak? A high-frequency evidence
Author: Zexi Sun (Goethe University)
Title: How does the stock market listen to Fedspeak? A high-frequency evidence
Abstract: The paper examines the effects of Federal Reserve communications on US equity prices using an intraday event-study analysis. Based on Bloomberg’s news feeds, I first construct a unique dataset that covers the key statements made by the members of the FOMC during press conferences, speeches and interviews. Then I propose a novel approach to quantitatively measure the language used in these statements, and to directly evaluate their impacts on the stock market. The result shows that, while the effects are highly nonlinear, these statements can explain a majority of the variation of returns in the narrow time window. Based on this approach, I also separate the “information effect” from the "pure policy effect" in Fedspeak, and find intuitive explanation of why these two opposite effects might arise from seemingly similar statements. Built upon cutting-edge machine learning algorithms, this approach is capable of accurately predicting the price change caused by any user-specified statement, and thus it can serve as an extremely convenient tool for policymakers to evaluate their communications on the financial market down to the word level.