Sentiment Analysis of Monetary Policy Communication: Part I
We find that the Federal Reserve influences other Central Banks, with an increasing magnitude after the Great Recession, proving its “leading role”. After the European debt crisis, the ECB gained relevance, whereas the BoE gradually reduced it. Finally, the BoJ seems to have a lower impact on other central banks’ communication.
According to Blinder et al., (2008), central bank communication refers to the quantity and quality of information made available to the public regarding the central bank’s monetary policy objectives and strategy; and its perception of the economic outlook as well as signals about future policy trajectory.
Naghdaliyev (2011) argued that the need to effectively manage expectations constitutes one of the key policy challenges of the post-2008 global financial crisis. He further emphasized that the provision of appropriate information to the public about central bank policy actions helps reduce rather than exacerbate economic uncertainties, thus making central bank communication a primary tool for expectations management.
Several channels are usually adopted for communication, including but not limited to monetary policy reports, minutes of monetary policy meetings, post-meeting briefings by central bank Governors, and speeches by Monetary Policy Committee (MPC) members. Recognizing the crucial roles of central bank communication in monetary policy effectiveness, many central banks have developed communication strategies to systematically disseminate their policy objectives and decisions to the public. For instance, the Governing Council of the European Central Bank holds its monetary policy meeting every six weeks, with the ECB President conducting a press conference after each meeting to explain its monetary policy decisions from a prepared statement, similarly, the FOMC (of the Federal Reserve -FED-) holds eight regularly scheduled meetings per year.
The communiques database
Here all the monetary policy documents communicated by governors and published by these central banks were analyzed using text mining tools and it was found that concerns relating to financial stability and economic uncertainties were quite high during the crisis. Following Loughran and McDonald (2011), we construct our own Central Bank Sentiment Indexes, applying text analysis to a novel dataset of 1066 documents, which comprises releases and speeches published by the Federal Reserve, the European Central Bank, the Bank of England, and the Bank of Japan between 2008 and 2020. The sources are the BIS Central bank speeches and for some cases, Central Bank’s websites.
Table 1 shows the total number of communiques for each central bank.Having processed the textual data into a form suitable for analysis, we examined the readability, and sentiments of the policy documents.
In Figure 1, we see that the central banks’ communication has decreased over the years, possibly suggesting that monetary policy transparency has not improved over the years. In addition, our results indicate that the words and terms driving the linguistic contents of the communique were influenced by the Bank’s policy objectives as well as the nature of shocks hitting the global and local economies.
Analyzing Figure 1, there may be a structural break in the series caused by the Great Recession, and the relative Financial Crisis, to prove this, we consider the average of Governor’s communique for each country, for example for the period 2008–2010, the BoE average communique published were 5, and increased considerably for the period 2011–2019 until 11 for 2019–2020, so for the case of the BoE we can say that they increased its transparency. On the contrary, for 2008–2010 the ECB published an average of 38 and the Fed 27 communique respectively, for 2011–2019, both institutions decreased the number 4 of communique, the ECB with an average of 32 and the Fed 16, for the last period, the average decreased too, ECB with an average of 27 and, the Fed with 12. Lastly, the BoJ follows the same declining pattern, for 2008–2010 with an average of 21, for 2011–2019 with 15, and 2020–2021 with only 11.
The sentiment index
This section aims to build and analyze a sentiment analysis model which will allow us to categorize monetary authorities based on their sentiments, that is whether they are positive, negative, and also the magnitude of their tone. This section employs text-mining techniques to analyze the communication strategy of the four world’s most important central banks during the period 2008–2021. Since the policy communique released after each meeting of the monetary policy committee (MPC) represents an important tool of central banks communication, we construct a corpus based on 1,066 policy communique: Bank of England (123), Federal Reserve (237), European Central Bank (478) and Bank of Japan (228).
To see if the texts may contain some useful insight to predict the Central Banks interest rate, we use the Loughran and McDonald Sentiment Word List1 to measure the sentiment of each statement. This dictionary contains several thousand words appearing in textual analysis in accounting and financial documents such as 10K, 10Q. It includes words in different forms, so stemming or lemmatizing should not be applied. We applied a simple technique to flip the sentiment for negation (e.g., can’t, isn’t, no, etc.).
Where 𝑁𝑝𝑜𝑠𝑡𝑖𝑣𝑒 denotes the number of words expressing positive sentiments, 𝑁𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒 is the number of words expressing negative sentiments, and NTotal is the total number of words in document. Any positive sentiment term used in a document takes a value superior to 1 while a negative sentiment term takes a value inferior to -1. It is implied in Eq. (1) that a value of MPSS > 0 indicates positive sentiments.
On Table 2, we see the MPSS results, it shows that the FED’s policy statements have recorded a less optimistic tone and more certainty to the negative impact of the crisis and economic phenomenon, while the tones of the policy statements published by the ECB and Bank of Japan (BoJ) have remained unchanged and with a more optimistic tone. Being Sir. King was the least optimistic governor (although it is important to consider the number of its documents of 23), while the most optimist was Mr. Draghi.
Sentiment analysis index and growth
Lastly, we will plot together the MPSS with the GDP growth. As shown in Figure 2, the issues relating to the impacts of economic shocks on each economy has remained an important source of concern to the monetary authorities, for example, we can show different peaks in the sample, in the first quarter of 2009 the FED, reached the most negative sentiment peak, while the BoE recorded its negative peaks in the second semester of 2016 during the economic recession; while the BoJ recorded its most negative peak during the last quarter of 2008, and finally, the ECB showed its one during the second quarter of 2009.
Our sentiment indicators mimic the general path of economic performance for each country’s economy. The lowest points of the sentiment index also coincide with the deteriorating economic outlook during the main economic recessions of the past two decades. For instance, the dip in the MPSS reflects the great recession (hat occurred between 2007 and 2009). For example, the lowest BoE’s MPSS, which was recorded in July 2016 reflects the Monetary Policy Committee’s perception regarding the potential risks to the attainment of its policy objectives concerning interest rates, and economic growth, which was 0.44%.
On the other hand, the BoE’s MPSS record two peaks in the first quarter of 2016, and the second quarter of 2019 with the MPC expressing its satisfaction with the performance of the economy, especially with regards to economic growth (of 0.54%). While the BoJ peaked in the third quarter of 2016, the Fed show its peak in the last quarter of 2016, last, the ECB recorded its peak one year later, at the end of 2017.
Finally, Table 3 provides some summary statistics over the sample period (2008–2021) and for two different sub-samples, namely the period of the crisis (2008–2010), the subsequent decade (2011- 2019), and the COVID-19 period (2020–2021).
Summarizing, studying the above three different periods for the MPSS and the GDP growth, we found that the BoE was the only monetary institution with a negative tone for all sub-samples, although England has one of the highest average GDP growth for 2011–2019 of 1.46%, while its MPSS average was of -0.239. Considering the Covid-19 period, we see that the BoJ is the only institution with a positive sentiment tone of 0.115, probably this is due that Asian countries having more experience handling massive epidemics.
The python code can be found on my personal portafolio at: https://vedia-jerezdaniel.github.io/exo, the different databases are not available due to permission constraints.
References
[1] Blinder, A. S., Ehrmann, M., Fratzscher, M., De Haan, J. and Jansen, D.J. (2008). “Central Bank Communication and Monetary Policy: A Survey of Theory and Evidence”. Journal of Economic Literature. 46(4), 910–945.
[2] Naghdaliyev, N. S. (2011). “Central Banks’ Communication in the Post-crisis Period”. The Harriman Institute, Columbia University.
Common Informative Acronyms
- CBSI -Central Bank Sentiment Index.
- FED -Federal Reserve.
- ECB -European Central Bank.
- BOE -Bank of England.
- BOJ -Bank of Japan.
- MPSS -Monetary policy sentiment score.
- GDP -Gross domestic product.
