Fake news spreading and sentiment of Italians during the first COVID-19 lockdown

The SEBCOV international study involves several countries including Italy. The project included an online survey, aimed at understanding the impact of COVID-19 on different social groups. The last optional question of the questionnaire is open-ended and asks for additional comments: it received a particularly high response rate. In these answers, it is possible to read personal experiences about the first lockdown (March, April 2020) given spontaneously, allowing the respondent to express an opinion without the researcher’s influence. Therefore, they represent a precious and original source of information about the impact of COVID-19 on Italians’ everyday life. We performed a sentiment analysis dividing the textual answers into three groups: positive, negative, or neutral content. We have also noticed that several answers contained fake news, thus we flagged them differently. Then, to evaluate the joint effect of other survey variables on the sentiment and on the fake news spreading, we fit two specific models. Thanks to a Bayesian regression model we identified the most significant variables to explain the respondents’ sentiment. Results of this model show that women tend to have a more positive attitude than men and that younger respondents more likely wrote negative sentences. The residence area have an influence on the sentiment : respondents living in the North of Italy have a higher probability to express a negative sentiment than those living in the South of Italy. This is probably due to the different distribution of the pandemic in the Italian territory. Moreover, who had to deal also with economic struggle during the COVID-19 lockdown presents a higher percentage of negative sentiment. Furthermore, we tried to deepen the fake news spreading phenomenon with a logistic model that emphasizes some interesting topics. We found that respondents who experienced an income loss during the lockdown more frequently reported a fake news as true in their comments at the end of the questionnaire. Men are more likely than women to write fake news in text and respondents who declare to get informed by traditional and institutional media write fake news in their comments less frequently. An apparently contradictory aspect is that those who self-judge able to detect fake news are also those who more easily include fake news in the final optional question. The final question of an elaborate international survey unexpectedly collected a lot of nonstructured data and proved to be a source of valuable information about the social struggle of Italians through the first COVID-19 lockdown. Our analysis provides a novel insight about Italians’ hurdles during lockdown, highlighting categories that were more affected, and contributes to understanding the determinants of fake news spreading during the pandemic.