What does text analysis tell us about central banks’ communication in times of COVID-19

July 20, 2020

Paola Priola, Giacomo Tizzanini, Piero Lorenzini

Over the last decade, communication by central banks has become increasingly important as a tool for managing monetary policy expectations. With text analysis techniques, it is possible to analyze the information embedded in speeches and press conferences and summarize it in an index, the Prometeia Central Bank Sentiment Index. From which important evidence emerges.

 

During his years as Chairman of the Federal Reserve, Alan Greenspan became the icon of a communication style that later took the name of "Greenspeak": he made wordy and ambiguous statements. "Since I’ve become a central banker," he said on one occasion, "I have learned to mumble with great incoherence. If I seem unduly clear to you, you must have misunderstood what I said". Until the 1980s, communicating à la Greenspan was common for central bankers, based on the idea that the effectiveness of monetary policy required an element of surprise for the markets.

Those years are long gone. Not only because of the financial crisis of 2008, which required constant efforts to stabilize the markets, but also because of the evolving role of communication for central banks in a context of higher economic fragility and uncertainty. In stark contrast to the previous tradition, central banks are now committed to transparency towards the public, providing details on decisions taken and on strategies for the future.

Recent literature has recognized the impact of central banks’ communication on the economy, as a result of both the evolution of communication and the increased frequency of publications (Dincer and Eichengreen, 2014), focusing not only on content but also on the economic sentiment embedded in the documents.

Unlike other economic indicators, however, sentiment indices require ad hoc methodologies to measure qualitative information (Grimaldi and Apel, 2012; Correa et al., 2017; Armelius et al., 2019; Picault and Renault, 2017). Using textual and lexical analysis techniques we built the Prometeia Central Bank Sentiment Index (CBSI), an indicator of central banks’ sentiment. In particular, we applied the approach introduced by Loughran and McDonald (2011), consisting in the identification of words with positive and negative connotations within a document.

The four central banks we examined represent the most influential economies in the world: the European Central Bank (ECB), the Federal Reserve (Fed), the Bank of Japan (BOJ) and the Bank of England (BOE). They are also ranked among the most transparent central banks (Dincer and Eichengreen, 2014), a necessary criterion for robust analysis.We examined speeches and press releases published on their respective institutional websites between 2000 and 2020. Due to the heterogeneity of the database in terms of topics and length, we adopted the methodology of Net Score Index (Birz and Lott, 2011). Indeed, releases are generally short and standardized documents, whereas the structure of the speeches varies depending on the event and the speaker.

In line with the literature and in order to obtain the Prometeia Central Bank Sentiment Index, we performed a pre-processing procedure on each document (i.e. cleaning and standardization), converting it into a corpus, that is, a structured and standardized text over which it is possible to perform statistical and computational analysis. Then, with tokenization, the text is split into single words with relative frequency (as well explained by Grun and Hornik, 2011), thus preparing the term-document matrix to calculate the index.

By applying the Loughan-McDonald dictionary, whose validity is corroborated by its vast use in the financial field (Armelius et al., 2019; Shapiro et al., 2020), we obtained the match of the words with positive (and negative) connotations of the dictionary in each document. The Sentiment Index is calculated on a monthly basis for each central bank. It takes values between -1 and +1 and provides a measure of the central bank's view (positive, neutral or negative) of the economic context, thus giving indications on the possible direction of subsequent monetary policy decisions.

 
Chart 1. Fed and ECB Prometeia CBSIs
 
Chart 2. BoJ and BoE Prometeia CBSIs
 

Following the outbreak of the pandemic in January 2020, the indicator shows that sentiment has declined sharply for all central banks considered.

In detail, the ECB, led by Christine Lagarde, hit its historical low twice in a row, first in March (-0.499) and then in April with -0.519. The Federal Reserve reached -0.318 in March, when it introduced extraordinary policies to mitigate the effect of the coronavirus on the economy. Instead, The Bank of England sentiment has been stable at -0.50 from the beginning of the Covid-19 crisis, whereas the Bank of Japan stood at -0.471 last May.

After the period of more intense lockdown, sentiment is now showing a slight improvement, demonstrating that the policies adopted – both monetary and fiscal – are perceived as an effective tool for the recovery of the global economy.

The pandemic has caused large-scale negative shocks, requiring central banks to intervene to support the economy of the most affected countries. The need to adopt effective economic policies required a resolute and assertive communication to reassure markets and institutions about the real possibility of recovery. As Ben Bernanke (Alan Greenspan's successor) once stated: "monetary policy is 98 percent talk and only two percent action".

 
Bibliography

[1] Armelius H., Bertsch C., Hull I. and Zhang X. (2019). “Spread the Word: International Spillovers from Central Bank Communication”, BIS Working Papers, No 824.
[2] Birz G. and Lott R. (2011). “The Effect of Macroeconomic News on Stock Returns: New Evidence from Newspaper Coverage”, Journal of Banking & Finance, 2011, vol. 35, issue 11, 2791-2800.
[3] Correa R., Garud K., Londono-Yarce J. and Mislang N. (2017). “Constructing a Dictionary for Financial Stability,” IFDP Notes. Washington: Board of Governors of the Federal Reserve System.
[4] Dincer N. and Eichengreen B. (2014). “Central Bank Transparency and Independence: Updates and New Measures”, International Journal of Central Banking, vol. 10(1), pages 189-259.
[5] Grimaldi B. and Apel M. (2012). “The Information Content of Central Bank Minutes”, Sveriges Riksbank Working Paper Series No. 261.
[6] Grun B. and Hornik K. (2011). “Topicmodels: An R Package for Fitting Topic Models”, Journal of Statistical Software, 40(13), 1-30.
[7] Loughran T. and McDonald B. (2011). “When Is a Liability Not a Liability? Textual Analysis, Dictionaries, and 10Ks”, Journal of Finance LXVI (1), 35-65.
[8] Picault M. and Renault T. (2017). “Words are not all created equal: a new measure of ECB communication”, Journal of International Money and Finance, No. 79, 36-156.
[9] Shapiro A. H., Moritz S. and Wilson D. (2020). “Measuring News Sentiment,” Federal Reserve Bank of San Francisco Working Paper 2017-01.


Read All