Analysing the impact of ECB Communication on Financial Markets: A Text Mining approach

Abstract

This thesis investigates how European Central Bank (ECB) communication made during the press conferences affects asset pricing and herding behaviour in the European financial market. My approach to analysing the ECB statement is through the lens of textual analysis. Specifically, the text in ECB press conferences is extracted by means of a Web Scraping procedure that divides ECB’s President transcript from the Q&A section. Then, Sentiment analysis is applied to both these two types of text corpus and used to construct a sentiment time series using different field-specific dictionaries, that span from the Monetary Policy to Financial Stability lexicon. I also created my own dictionary merging different lexicons found from previous studies. These time series are then used as explanatory variables to compare to the Euro Stoxx 50 time series. Evidence suggests that my sentiment time series reached the highest explanatory power in predicting Euro Stoxx 50 market realizations. Additionally, to assess the evidence of whether herding behaviour of investors occurs around positive and negative spikes of the sentiment time series, I applied the Belgacem and Lahiani (2013) methodology to different European sectors. I found that those spikes do help in detecting herding behaviour in the European financial markets.