To quantify the content of ECB introductory statements, we consider all sentences pronounced by the ECB President. Then we classify them manually into seven categories:
The first three categories are grouped into a topic labeled Monetary Policy (MP) and refer to the ECB Governing council monetary policy decisions, without considering references to past decisions. This topic also includes references to the short and medium term views of the Governing council on the expected path of monetary policies. The next three categories are grouped into a topic labeled Economic Outlook (EC) and focus on policy makers' description of the current economic situation and views on the future economic outlook. The last category None groups sentences not directly relevant to either monetary policy decisions or the Governing council economic outlook. This category also includes sentences presenting data that have already been released before ECB statement without any forward-looking statement nor additional information.
Then we decompose all sentences into group of words (from one word to a maximum of 10 words for a single group). And we simply count the number of times this group of words is classified in each of the seven categories and we compute the probability that it belongs to the category MP and EC with a given inclination (dovish, neutral, hawkish for MP -- positive, negative, neutral for EC). The probabilites for each group of words are available for download here.
Then we aggregate the probabilities for each categories and inclinations at a press conference level to obtain the probabilities detailled in the data section. An introductory statement final score is the difference between :
© CBCI 2017