Extracting the news(worthy) from the noise

Global central banks laboured under increased scrutiny following the Global Financial Crisis (GFC) of 2008. Having embarked on extraordinary monetary policy experiments, it followed that considering shifts in prevailing central bank policy, and extrapolating any likely effects became an expedient tactic.

In the spirit of such goals, and given the European Central Bank (ECB) recently made public a cache of two decades worth of speeches, we introduce the reader to a low-level text analysis using commonly available Natural Language Processing (NLP) tools.

We highlight characteristics of the text data; reveal extractable features of central bank vernacular; identify main themes and their evolution; apply a sentiment extraction engine to measure an aggregate disposition of policy makers; and, finally, test the foretelling ability of sentiment vis-à-vis changes in monetary policy.