The role of textual data in finance: methodological issues and empirical evidence

This thesis investigates the role of textual data in the financial field. Textual data fall into the more extensive category of alternative data. These types of data, such as reviews, blog post, tweet, are constantly growing, and this reinforces the importance in several domains. The thesis explores different applications of textual data in finance to highlight how it is possible to use this type of data and how this implementation can add value to financial analysis. The first application concerns the use of a lexicon-based approach in the credit scoring model. The second application proposes a causality detection between financial and sentiment data using an information-theoretic measure, the transfer entropy. The last application concerns the use of sentiment analysis in a network model, called BGVAR, to analyze the financial impact of the Covid-19 Pandemic. Overall, this thesis shows that combining textual data with traditional financial data can lead to a more insightful knowledge and, therefore, to a more in-depth analysis, allowing for a broader understanding of economic events and financial relationships among economic entities of any kind.