Statistical analysis of Twitter data to assess the relationship between sentiment towards Italy and stock exchange trend

Twitter is extensively used to share news, comments, opinions and reactions to events. During the recent Coronavirus disease 2019 (COVID-19) pandemic, Italy was among the first countries in Europe to be severely hit by the outbreak and to establish measures such as lockdown and stay-at-home orders. We hypothesize that this might have resulted in a country reputation damage. We investigate changes in opinions about Italy before and after the COVID-19 outbreak using sentiment analysis on a large dataset including over 240,000 posts extracted from Twitter. First, we apply different lexicons-based methods and show a relevant change from a positive to a negative sentiment towards Italy corresponding to the date of the first established case of COVID-19 in the country. Next, we show a significant positive correlation between sentiment about Italy and values of the FTSE-MIB index, the main index for the Italian Stock Exchange. These findings suggest that sentiment might be useful to early detect changes in stock exchange values. Finally, we compare performances of two widely used supervised classification models, Naïve Bayes and Support Vector Machine, in the classification of tweets posted before and after the outbreak into a positive or negative class. Our results might be of help to gain insights into the relationship between sentiment measured using social media data and potential economic repercussions for a country.