In coll. with: Shantanu Banerjee, Michiel De Pooter, Brad Strum and Cait Walsh
Abstract : Using articles in the financial press, we use natural language processing to construct an index of the perceived sentiment of FOMC communications around FOMC meetings. Using changes in our index around each FOMC meeting between May 1999 to June 2017 as a measure of monetary policy surprises, we find that changes in market-perceived FOMC sentiment explain movements in asset prices during narrow FOMC event windows. We also document that most of this explanatory power appears to be driven by the sample prior to the zero lower bound period. We compare our results with sentiment indexes constructed using other methodologies in the literature and find that our index has higher explanatory power for movements in various financial asset prices around FOMC meetings, even when we control for financial market-based measures of monetary policy surprises. Through a series of robustness checks, we show that our results are robust to several alternative specifications of our index.