13:20 | 14:00
Keywords defining the session:
- Sentiment Analysis
- Monetary policy
Takeaway points of the session:
- We measure and monitor Central Banks’ communication strategy using large databases of text from their monetary policy reports and press releases thanks to the use of machine learning algorithms for natural language processing and topic modelling
- We found that the NLP techniques are an adequate tool to the analysis of monetary policy in Central Banks since they help us to understand the main challenges and the evolution of Monetary Policy.
Big Data and Data Science techniques allows us to measure and analyze text using natural language processing methodology, also known as text mining or computational linguistics. The information included in the media, blogs, economic and financial reports, etc. in the form of text could fully complement and improve our structured databases traditionally used in economic research. Thus, using statistical techniques and computational tools, we quantify text extracting meaning from letters, that is, we convert text into data. This novel approach, which permits to complement and combine traditional economic tools with new emerging tools thanks to the use of Big Data, has plenty of applications with a huge potential for economic research.
We focus on the monetary policy analysis through the measurement and monitorization of the Central Banks’ communication strategy. This communication strategy refers to the information that the Central Banks release about the current economic situation, its current and future policy intentions together with the expected path for future monetary policy decisions. It has become an important tool for understanding monetary policy decision. Moreover, it is also a complementary tool for Central Banks to manage expectations about the future direction of the monetary policy.
We measure and monitor Central Banks’ communication strategy using large databases of text from their monetary policy reports and press releases thanks to the use of machine learning algorithms for natural language processing and topic modelling. Furthermore, we exploit the emotional dimension through sentiment analysis to extract semantics of opinion words and sentences in the text and analyse sentiment patterns across topics using the Lexicon approach.
Previous literature has used these computational linguistics models to analyse the Federal Reserve communication transparency strategy (Hansen et al, 2014) as well as the effects of this communication strategy on real economic variables (Hansen et al, 2015). From a methodological point of view, they do it using Latent Dirichlet Allocation (LDA)1 that estimates what fraction of time each speaker in each section of each meeting spends on a variety of topics. Our main contribution to this research area is to measure Central Bank communication strategy for an Emerging Country like Turkey and to go one step ahead in the computational approach to incorporate dynamics and topic evolution to this analysis using Dynamic Topic Models (DTM), which is a particular case of Structural Topic Models (STM) (Roberts et al. 2013).
Beyond the NPL techniques, this working paper complements the analysis of Central Bank communication policies and its effects on financial markets and real variables. The NPL analysis described in the first part of this paper will help us to understand deeply how the Emerging Markets’ Central Banks talk. Particularly, we detail the most important included topics in the communication documents, what is their sentiment in the text when they are commented and how they interact over time. In the second part, we analyse how the monetary policy communication impacts on financial markets and the real economy, thus complementing the literature of the Central Bank communication policies.