With 2 fellow colleagues, we replicate and extend the work of Amaya and Filbien (2015) who examined the similarity of communication at European Central Bank press conferences and their impact on financial markets. To extend their work, we investigate non-linear effects of inflation and time on similarity measure as well as the non-linear effect of similarity on absolute cumulative abnormal returns. Despite slight difference in the data source, we first show that from January 1999 to December 2013, our results are qualitatively similar to their findings. Most importantly, we evidence significant non-linear effects (U-inverted) while extending the sample to December 2022. Note that a 10 pages research paper has been written based on this work.
We analyze and assess the similarity of ECB conference speeches over the period 1999 to 2022. Overall, this study aims to provide a comprehensive overview of the similarities in the communication practices of the ECB, and to shed light on the role that communication plays in affecting market outcomes. First, (i) we compute similarity measure by using the Jaccard similarity score between two subsequent speech. Then, (ii) we turn into investigating any linear trend in the computed similarity measure. The purpose of this approach is to test if the speech of the European Central Bank becomes increasingly similar over time. To measure european stock market’s reaction to ECB communication announcements, (iii) we conduct an event study which allows us to obtain Cumulative Abnormal Return (CAR) surrounding ECB announcements. Afterwards, (iv) we compute an informational content (pessimism) of ECB communication; which we interact with our similarity measure to see how informational content condition on the degree of similarity affects the market outcome. To extend the paper by Amaya and Filbien (2015) , we first account for non linear effects of inflation and time on similarity measure. Second, we investigate the nonlinear effect of similarity on the absolute cumulative abnormal returns. Overall, we find significant non linear effects (U-inverted) of these variables on the respective response variables.
We argue that the major changes that occurred at the ECB board between the 2013–2022 window and other major economic events might in part explain the significant non linearities evidenced in this work. Future research could explore the specific factors driving these linearities or compare the results found for the ECB with those of other central banks. A limitation of this study is the limited access to standardized and comprehensive data sets. Not only the authors did not discuss how their data were standardized but also some of the provided data sources were inaccessible. Hence, we are aware this work could have been better if we had access on a comprehensive and coherent data sets beforehand.
Every step of this work is very well described in the attached notebook.