Three essays on the economic perspectives of false information

This thesis covers the debate about the political outcomes of false information online. The dissertation is composed of three articles. The first article sets the theoretical basis for the empirical analysis and proposes a cross-disciplinary classification of studies in the field. I identify five main groups of studies: the first group collects studies examining the spreading patterns of false information online; the second investigates the cognitive mechanisms underlying individuals’ information processing, which may lead to biased beliefs; the third covers analyses testing the efficacy of debunking in adjusting incorrect beliefs; the fourth encompasses studies aiming to identify successful automatic detection algorithms for false information; the last includes studies exploring the political outcomes of false information. The second article investigates how false information relates to the performances of the main parties across Italian provinces in the context of the 2018 political elections. I draw on geotagged Twitter data to build a measure of exposure to misleading contents and then match this information with other province-level data. Results suggest that false information played a role in the Italian political debate. I find that the index is positively and significantly associated with the performance of the Five Star Movement and negatively associated with the performance of the Democratic Party when the analysis is restricted to political disinformation or a subsample of pro-disinformation tweets. The third article investigates the impact of false information online on the 2019 European election in Italy. The first part of the paper carries out a cross-sectional analysis exploring how the exposure to false information on Twitter relates to the electoral outcomes. Results show a positive relationship with the performances of the Five Star Movement and the Democratic Party. In the second part of the paper, a First Difference analysis is performed exploiting data collected in the occasion of the 2018 political election, which confirms the association found for the Five Star Movement but not for the Democratic Party. I conclude arguing that these results are suggestive of the presence of online polarization and consistent with this literature.