The Response of Exchange Rate to Central Bank Communication: A Textual Analysis Approach

Abstract

The study explores how central bank communication affects the movement of exchange rates. The paper constructs sentiment by employing textual information from the European Central Bank press conferences (ECB PCs). Using high-frequency data, we find that a more positive/negative sentiment is associated with appreciation/depreciation of euro and decreasing/increasing exchange rate volatility. The results are robust to the inclusion of relevant control variables: the changes in key policy interest rate, textual characteristics of ECB PCs, and macro fundamentals. We apply a machine learning technique to decompose central bank communication into a variety of topics. The results indicate that EUR-USD responds most strongly to topics of economic growth and central bank liquidity provision. We provide evidence that central bank sentiment contains forward-looking views about future policy rates and exchange rates. A risk-based channel is only significant during the unconventional monetary policy period.

Keywords: Central bank communication, exchange rate, sentiment, LDA, textual analysis

JEL Classification: G11, G12, E52, E58